Low-temperature tunneling of CH3 quantum

rotor in van der Waals solids

Cite as: Low Temp. Phys. 45, 427 (2019); https://doi.org/10.1063/1.5093524
Published Online: 03 April 2019

Nikolas P. Benetis, Ilia A. Zelenetckii, and Yurij A. Dmitriev

ARTICLES YOU MAY BE INTERESTED IN

Critical temperature of the superconducting transition of individual phases of multiphase

bismuth cuprates after cooling in a magnetic field to a temperature of 77 K
Low Temperature Physics 45, 386 (2019); https://doi.org/10.1063/1.5093517

Influence of two-level defects on the transmission of a microwave signal in an open coplanar

waveguide
Low Temperature Physics 45, 395 (2019); https://doi.org/10.1063/1.5093518

High pressure-induced relaxation of electrical resistance in weakly doped НоBa2Cu3O7–x

single crystals
Low Temperature Physics 45, 465 (2019); https://doi.org/10.1063/1.5093527

Low Temp. Phys. 45, 427 (2019); https://doi.org/10.1063/1.5093524 45, 427

© 2019 Author(s).
 Low Temperature ARTICLE scitation.org/journal/ltp
Physics

Low-temperature tunneling of CH3 quantum rotor

in van der Waals solids
Cite as: Fiz. Nizk. Temp. 45, 495–510 (April 2019); doi: 10.1063/1.5093524
Submitted: 21 February 2019 View Online Export Citation CrossMark

Nikolas P. Benetis,1,a) Ilia A. Zelenetckii,2 and Yurij A. Dmitriev3

AFFILIATIONS
1
Department of Environmental Engineering and Antipollution Control, Technological Educational Institute of Western
Macedonia (TEI), Kila 50 100 Kozani, Greece
2
Department of System Analysis and Control, Institute of Computer Science and Technology, Peter the Great
St. Petersburg Polytechnic University, 29 Politekhnicheskaya ul., St. Petersburg 195251, Russia
3
Ioffe Institute, 26 Politekhnicheskaya ul., St. Petersburg 194021, Russia

a)
E-mail: niben@eie.gr

ABSTRACT
Motional quantum effects of tunneling methyl radical isolated in solid gases as they appear on experimental electron paramagnetic
resonance (EPR) spectra are examined. Obtained analytical expressions of the tunneling frequency for methyl rotor/torsional-oscillator uti-
lizing localized Hermite polynomials are compared to full numerical computations and tested against experimental EPR lineshape simula-
tions. In particular, the X-band of methyl radical was displaying partial anisotropy averaging even at lowest temperatures. EPR lineshape
simulations involving rotational dynamics were applied for the accurate determination of the potential barrier and the tunneling frequency.
Tunneling frequency, as the splitting between the A and E torsional levels by the presence of a periodic C3 model potential with periodic
boundary conditions, was computed and related to the EPR-lineshape alteration. The corresponding C2 rotary tunneling about the in-plane
axes of methyl was also studied while both the C2 and C3 rotations were compared with the rotation of deuteriated methyl radical.

Published under license by AIP Publishing. https://doi.org/10.1063/1.5093524

1 INTRODUCTION The molecules of cryocrystals are held together by weak van

A vast number of processes in solid state are realized der Waals forces and are among the simplest solids to test ab initio
through potential-barrier hindering dynamics of hosted neutral theoretical approaches. However, even in these solids, composed of
or charged particles/probes. For example, at low temperatures, the most weakly bounded particles, the interactions contributing to
only way for a particle to overcome a potential barrier is by the energy of the matrix particles may reach rather high values.
quantum tunneling which becomes crucially enhanced for light For example, the central attraction energy of a matrix molecule in
atoms and molecules. Among those, methyl radical (CH3 and solid CO2 is of the order of 10787 cal/mole1 which is equivalent to
its isotope analogues) is considered as quantum rotator because 5428.3 K. The rotational tunneling and librations of the matrix iso-
of its small inertia. The radical may be stabilized in chemically lated CH3 radical are, thus, the major mechanisms of the radical
inert matrices of which cryocrystals are of particular interest. reorientation at the cryogenic temperatures. The reorientational
Monoatomic examples of host matrices are Ne, Ar, Kr, Xe, while motion of CH3 in molecular cryocrystals was found to be complex
molecular-gas host matrices of general cryogenic interest are in nature and included fast rotation around the molecular C3 axis,
hydrogen H2, nitrogen N2, oxygen O2, carbon monoxide CO, fast librations around the in-plane C2 axis, and slow tunneling tum-
carbon dioxide CO2, nitrous oxide N2O and methane CH4. The bling about C2 axis.2–4 The observed temperature dependence of
rotational behavior of methyl radical embedded in these matrices methyl radical motion in these previous studies clearly evidences
is very sensitive to the state and dynamics of the surrounding specific changes occurring in the orientation dynamics of the
matrix molecules. Therefore, electron paramagnetic resonance matrix molecules. The results are in line with a very recent study5
(EPR) which provides information on such phenomena is a good by Krainyukova and Kuchta who reported new insight on the
tool to study the solid state behavior of such embedded radicals at molecular structure and dynamics of solid CO2. These authors per-
very low temperature. formed a high-energy electron diffraction study on solid carbon

Low Temp. Phys. 45, 000000 (2019); doi: 10.1063/1.5093524 45, 000000-427
Published under license by AIP Publishing.
 Low Temperature ARTICLE scitation.org/journal/ltp
Physics

dioxide films in the temperature range 15–87 K and found hopping become secondary for temperature higher than 50 K. In this
precession of molecules instead of simple small-angle librations estimation the torsional energies Eν = 3(BV3)1/2(ν + 1/2) were used,
that should not exceed 5–6 degrees. The relevant maximal angle vide infra, for rotational energy B ca 7 K and potential barriers
deviations turned out to be as big as ∼30 deg at the lowest tempera- V3 = 1000 K. Notice also that barriers over 1200 K studied by
tures (∼15 K) and started to decrease with rise in temperature at Stejskal and Gutowsky10 were exceedingly larger than the one used
∼45 K. This temperature point almost exactly coincides with the in the latter estimation and the ones of interest in the present work.
onset of the anisotropic C2 rotation we observed in our methyl Following the above discussion, we simplify our calculations
radical EPR study in solid CO2.4 It is worth noting that, earlier, the by not incorporating higher torsional levels than the ground level
rank-4 orientational order parameter η4 was found to decrease as also explained by Yamada et al.11 Further details of this
rapidly in pure solid CO2 at temperatures above 45 K.6 This η4 approximation are discussed in the section about the theory of
parameter determines the correlative and rotational-unharmonic the tunneling-methyl radical. The general quantum mechanical
effects in the librational subsystem. In fact, EPR studies of CH3 sta- problem of tunneling for hindering potentials of C3 and C2 sym-
bilized in solid Kr films discovered formation of a highly disordered metries are also treated easily in the present work within a
matrix structure for samples condensed from the gas phase at numerical formalism. The applied method is also applicable for
4.2 K7 as well as 16 K.8 The structure accumulated the majority of potentials of arbitrary complexity provided that its matrix ele-
trapped radicals which survived annealing to 31 K.8 The radicals in ments in the simple Bloch-type basis set are available.
the disordered structure yielded very broad hyperfine (hf ) lines The article is organized as follows: In Sec. 2, the experimental
compared to those at the centers in the symmetric substitutional matrix shift of the EPR parameters and the anisotropy of methyl,
sites of the regular Kr lattice. The broadening was found to origi- CH3, radical isolated in four different matrices consisting of spheri-
nate from the distribution of the radical g-factor due to the matrix cal molecules are analyzed in relation to the radical tunneling fre-
effect. An attempt of assessing the disorder parameter was pre- quencies at cryogenic temperatures. The role of librations in the
sented in Ref. 3. averaging of the parameter anisotropies is also considered for
The present study is aimed at elucidating matrix effects in slightly higher temperatures where quantum and classical motion
rotationally averaged methyl radical parameters and correlating coexist. The results are utilized in Secs. 3 and 4 to theoretically
tunneling rates to barriers of rotation for CH3 in van der Waals address the rotational rates and the libration amplitudes of the
solids. The results are essential for converting the radical to a probe trapped methyl radicals even in matrices consisting of linear mole-
to determine low-temperature structures and dynamics processes cules. In particular, the tunneling frequencies of methyl radical
occurring in cryocrystals. against periodic C3 or C2 hindering potentials are computed both
An estimation of the tunneling particle mobility is possible analytically and numerically exploiting the experimental EPR
through the tunneling frequency9 which can be used identically as results. The EPR results correspond, respectively, to parallel or per-
the torsional splitting of the localized degenerate vibrators’ ground pendicular rotary tunneling motions of methyl radical in different
level. The torsional splitting is thus per definition a temperature solid gas matrices, concerning exclusively very low-temperature
independent quantity as it requires an isolated quantum system. quantum conditions. The main task is to show that a wealth of per-
Experimentally, the torsional splitting is observable at lowest tem- ceptible orientational motions is available for a 3D quantum
perature while the definition of the tunneling frequency νt is given rotator even at cryogenic temperatures by tunneling through barri-
differently by Stejskal and Gutowsky:10 ers. Finally, after a short concluding Sec. 5, an Appendix demon-
strating the mathematical details of the computations is included.
1 X E
i =RT
vt ¼ vi e : (1)
Q i
2. HINDERED MOLECULAR ROTATION EFFECT ON THE
The average energies E
i of each originally two-fold degenerate EPR PARAMETERS
localized torsional level, i, (C2 potential was assumed in that work), The EPR lineshape of CH3 trapped in van der Waals (vdW)
in the above Eq. (1) weights the corresponding torsional splitting solids at cryogenic temperatures demonstrates significant variabil-
νi. The partition function Q, comprising the sum of the Boltzmann ity. It changes from nearly symmetric for methyl radical in solid
exponentials, is the normalization factor. The Boltzmann statistics p-H2,12 to asymmetric spectrum with split anisotropic lines and
used to obtain the populations of the torsional levels thus renders additional satellite transitions of non-rotating methyls, in cold solids
the above “tunneling frequency” νt as temperature-dependent consisting of the linear CO2, N2O molecules. The above two sets of
function. main and additional transitions are characterized by either isotropic
There is, however, a fundamental drawback for the usefulness or axial spin-Hamiltonian vs. rhombic spin-Hamiltonian, respec-
of the definition in Eq. (1) in quantum systems; in particular, for tively,13,14 depending on the actual dispersion forces between matrix
large potentials of the order of 1000 K and higher. The difference and CH3 radicals.15 The experimental anisotropy variation of the
in the energies of the torsional levels is too large to give any signifi- EPR spectra of methyl radical in the present work was related to
cant population to any other levels than the ground level. increasing rotational rate of the radical with temperature increase.
In many other systems of tunneling methyl radical, similar to The variation of the radical-matrix interaction influences the
our case, a simple estimation gives the observable effect from the matrix shifts of the EPR parameters and changes the radical
higher torsional levels for temperatures higher than 250 K, while dynamics governing the parameter averaging. Moreover, it is not
experimentally, the importance of the motional quantum effects possible to determine the exact static anisotropy of the g and A

Low Temp. Phys. 45, 000000 (2019); doi: 10.1063/1.5093524 45, 000000-428
Published under license by AIP Publishing.
 Low Temperature ARTICLE scitation.org/journal/ltp
Physics

tensors experimentally, since methyl tunnels (rotates) even at the Below the liquid helium temperature methyl radicals are in the
lowest experimental temperatures. ground rotation A-symmetry state yielding axially symmetric g and
Radical species isolated in a matrix cage are allowed to A hf tensors with parallel component coinciding with the highest C3
perform orientational and/or oscillatory motions. In case of a weak symmetry axis of the radical.13 The spectrum averaging is achieved
matrix coupling, the trapped molecules rotate almost freely in the by additional fast rotation about the in-plane axes (perpendicular
host cage, yielding nearly isotropic EPR parameters with small rotation). The rotation is characterized by tunneling frequency esti-
matrix shifts. Free rotation quantum effects at 4.2 K were observed mated as the reciprocal of the rotation correlation time, τcorr, consid-
early by Foner et al.16 in the EPR experiments of NH2 radical. ered as measure of rotation hindrance. One of the aims of the
These effects were further explained by McConnel17 who also present study is to verify the exclusive effect of the Pauli interaction
foresaw the liquid He 1:1:1:1 methyl quartet. Extending the same on the spectrum anisotropy and obtain information about the CH3
basic considerations, high-resolution EPR spectra with obvious rotation dynamics by measuring and correlating τcorr to the radical–
Pauli excluded EPR transitions of methyl hydrogen isotopomers matrix interaction.
were observed experimentally in cryogenic Ar matrix and inter- First, we address the case of methyl radical in solid Ar; at
preted accordingly.11 Matrix effects of CH3 in solid gases of spheri- 4.2 K, the spectrum of solid Ar is composed of four hf lines with
cal particles, in particular, spectrum anisotropy, were studied different amplitudes, A, and widths, ΔH.18 The line intensities, esti-
semiquantitatively in Ref. 12, setting the following empirical coeffi- mated as the products A(ΔH)2 are nearly equal yielding the well-
cient, Arel, as its measure: known 1:1:1:1 low-temperature intensity ratio. The different widths
and amplitudes of the four components are due to the residual
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffirffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
A0 max A0 max anisotropy after partial rotational averaging.pIn
ffiffiffiffi Fig. 1(a), the recip-
Arel ¼ 1 þ1 : (2) rocal square roots of the hf amplitudes, 1= A are proportional to
A0 min A0 min
the line width12 that may be approximated by the following qua-
dratic in mF formula:19
In Eq. (2), A0 max and A0 min are the amplitudes of the strongest
and weakest hyperfine (hf ) components of the EPR quartet. It was
shown that the anisotropy was governed mainly by short-range 1
pffiffiffiffi / ΔH ¼ a þ bmF þ cm2F : (3)
Pauli exclusion forces while the contribution of the van der Waals A
attraction was negligible. Varying amplitudes and widths of the
different hf components indicate residual magnetic anisotropy due Here, mF is the projection of the coupled nuclear spin representa-
to incomplete rotational averaging of the spectral parameters. tion F = I1 + I2 + I3 of the three protons.

FIG. 1. EasySpin simulations of the

EPR spectra for matrix isolated CH3 at
4.2 K in solid gases: in Ar (a), Kr (b),
Ne (c), and para-H2 (d). The reciprocal
square root amplitudes of the hf com-
ponents are plotted against the projec-
tion mF of the three coupled-proton
spin representation F. The blue dashed
line calculated using Eq. (3) fits the
experimental data presented in black
asterisks. The experimental data were
taken from our earlier study.12 The
EasySpin simulation results indicated
by red open circles are obtained with
EPR parameters listed in the text.

Low Temp. Phys. 45, 000000 (2019); doi: 10.1063/1.5093524 45, 000000-429
Published under license by AIP Publishing.
 Low Temperature ARTICLE scitation.org/journal/ltp
Physics

The procedure of obtaining the CH3 EPR parameters and the simulated hf components are 0.0248, 0.0163, 0.0141, 0.0172 mT
rotation correlation times was divided into two steps as follows: counting from low to high fields and match well with experimen-
First, the experimental reciprocal square root amplitudes of the hf tal widths18 0.0229, 0.0183, 0.0136, 0.0204 mT.
components in Figs. 1(a)–1(d) were fitted by adjusting the coeffi- To make a complete consideration, we tried another approach
cients a, b and c of Eq. (3) assuming fast-motion regime conditions. by fixing Δg = −3.7⋅10–4 equal to that of the g-tensor anisotropy for
The Origin 6.1 routine was used to determine the polynomial CH3 in solid N213 and changing the A-tensor anisotropy. Figure 2
coefficients minimizing the standard deviation (SD) of the best summarizes the EasySpin obtained simulations.
curve fit through the experimental data. The upper spectrum is simulated using the EPR parameters
The EPR parameters were obtained from the above fitting pro- listed above and fits well with the CH3/Ar experimental data shown
cedure using reasonable estimates of the magnetic parameters as in Fig. 1(a). The next four simulated spectra show changes in the
input in EPR spectral simulations by the EasySpin software. This EPR lineshape with fixed Δg and gradual increase of ΔA. The simu-
procedure accounts for Lorentzian and/or Gaussian profiles allow- lations begin to resemble (however, poorly) the experimental
ing accurate comparison between simulated and experimental line results when a large value of the hf-tensor anisotropy, ΔA, is
widths. It is a direct method, compared to using the standard involved, not meeting the physics of the system. Indeed, the radical
expressions for the coefficients a, b, and c of Eq. (3), see Ref. 12
and references therein.

2.1. Determination of methyl radical tunneling

frequency
The effect of motional averaging of the EPR parameters
anisotropy of methyl radical isolated in matrices of spherically sym-
metrical particles is considered in this work. The perpendicular
rotation of methyl in these matrices is not hindered compared to
matrices of linear molecules where strong hindrance is the case.20
For CH3 isolated in solid Ar, the approximate curve obtained
in a least-squares fit of the experimental data is plotted in Fig. 1(a).
To obtain the correlation time of the rotation, the simulation of the
EPR spectrum with the EasySpin software21 was performed in the
fast-motion regime. The physical parameters of solid Ar and solid
N2 are close as anticipated by the full mutual solubility of the two
components. Therefore, matrix shifts of the EPR parameters of
methyl radicals trapped in Ar and N2 solids are expected to be
close. Referring to the hf couplings, they are, indeed, nearly
equal: –2.313(5) mT for CH3 in Ar compared to −2.317(5) mT for
CH3 in N2 matrix.13 The isotropic g-factor, however, shows a
noticeable difference, 2.002322(56) and 2.00250(12) for CH3 in Ar
and N2 solids, respectively. The g-factor matrix shift of methyl is
due to admixture of the unpaired electron p-orbital of the radical
with the frontier orbitals of the neighboring matrix particles. This
effect is similar to the matrix isolated atomic hydrogen, whose
g-factor matrix shift is obtained by admixture of its electron wave
function with the p-orbitals of the nearest matrix particles after
the required orthogonalization.22
The principal quantum numbers of the orbitals of the outer
(valence) Ar electrons are larger compared to the N2 molecule and
would lead to an increased g-factor shift. To verify this possibility,
we performed test simulations keeping the A-tensor anisotropy, FIG. 2. EPR lineshape of CH3 in solid Ar simulated using different values of
ΔA = A|| – A⊥, equal to that of CH3 in solid N2, ΔA = 0.098 mT, the g- and A-tensor anisotropies. The microwave resonance frequency was fres
= 9.4003 GHz. The isotropic components of the tensors were set equal to those
and changing simultaneously the g-tensor anisotropy, gΔ = g|| – g⊥,
obtained experimentally: giso = 2.002322, Aiso = −2.313 mT. (a) ΔA ≡ A|| − A⊥
the individual line width, ΔH, and the τcorr to obtain the best possi- = 0.098 mT, equal to the value measured for CH3 in solid N2, Δg ≡ g|| − g⊥ =
ble curve fitting. The isotropic parts of the A- and g-tensors were set −1.73⋅10–4, τcorr = 24 ns; (b)–(e) Δg = −3.7⋅10–4, equal to the value measured
equal to those of CH3 in solid Ar: Aiso = −2.313 mT, giso = 2.002322. for CH3 in solid N2, and τcorr = 20 ns for all four simulations, whereas the
The individual line width was isotropic and the lineshape A-tensor anisotropy ΔA increases gradually as 0.3, 0.5, 1.0, 1.2 mT, respec-
was Lorentzian. The best fit was obtained with Δg = −1.73⋅10–4, tively. Figure 1 evidences that the simulation (a) is very close to the experimen-
tally measured amplitude ratios of the hf components of the spectrum recorded
ΔH = 0.0086 mT, τcorr = 24 ns. In Fig. 1(a), the results of the simu- in the CH3–Ar experiments.
lations fit well to the approximate curve. The peak-to-peak widths of

Low Temp. Phys. 45, 000000 (2019); doi: 10.1063/1.5093524 45, 000000-430
Published under license by AIP Publishing.
 Low Temperature ARTICLE scitation.org/journal/ltp
Physics

reorientation about the in-plane axes is considered as a combina- As a result, the following values were obtained: ΔA = 0.138 mT,
tion of very fast small-angle librations with the frequency of the Δg = 1.13⋅10–4, τcorr = 17 ns. The opposite sign of Δg compared to
order of 1012 Hz and much slower full rotation with frequencies CH3 in Ar stems from the greater matrix shift of the g⊥ component
less than 109 Hz.20,24,25 The anisotropies Δg and ΔA, used in the compared to the g|| component reported earlier.13 Comparing giso in
spectral simulations, result from the rigid radical parameters, par- the two noble gas matrices, it is concluded that Kr matrix affects the
tially averaged by radical librations. Previously, the averaged g-factor far more efficiently than Ar matrix. It is also worth to pay
anisotropies were shown to gradually decrease with decreasing attention at the hyperfine tensor anisotropy yielded by the simula-
eccentricity of the matrix molecules in the order, CO2, N2O, CO, tion. The value 0.138 mT lies between the anisotropies for CH3 in
N2.20 Accordingly, ΔA for Ar matrix should be equal or somewhat N2O and CO2 matrices, being among the largest hf anisotropies for
smaller than N2 matrix. matrix isolated CH3 in solid gases. The radical libration angles in the
The case of Kr matrix is of particular interest. Indeed, the substitutional position of the regular solid Kr lattice are expected to
amplitude sequence of the EPR spectrum of CH3 in solid Kr matrix be larger than in Ar and N2 matrices because of the larger lattice
is a mirror image of the CH3 spectrum in practically all the other constant of 6.13 Å in Kr, compared to 5.31 and 5.65 Å in Ar and N2,
solid gas matrices, Ne, Ar, N2, CO, N2O, CO2.13,18,19 This “mirror” respectively. Large-angle librations should, however, tend to decrease
effect originates from the opposite sign of g-tensor anisotropy in Kr the hf anisotropy.20 An explanation for this inconsistency could
compared to Ar matrix,13,18 modifying the linear, cross relaxation be the local rearrangement of the strained crystal lattice of
term of Eq. (3). Another feature of the CH3–Kr system is the largest quench-condensed Kr. The rearrangement of the cubic lattice to
matrix shift of the g-tensor compared to CH3 in other matrices,13 axially symmetric hexagonal affects the axial anisotropy of the EPR
including solid gases, zeolites, vycor glass, beryl, methane hydrate, parameters. This process was suggested previously when consider-
and feldspar. On the contrary, the CH3–Kr hf coupling does not ing axially symmetrical EPR of H-atoms trapped in quench-
seem to demonstrate such striking matrix shift characteristics: the hf condensed Kr26 and saturation peculiarities of CH3 radicals in the
coupling anisotropy has the same sign as the CH3–Ar system13 and same matrix.3
the hf coupling matrix shift is of moderate value.20 The EPR transi- Figure 1(c) shows the EPR lineshape simulation data of CH3
tions recorded in the CH3–Kr experiments are somewhat broader19 trapped in solid Ne. The best lineshape fit was obtained with
compared to those of the CH3–Ar system. The extra broadening is g- and hf A-tensor anisotropies close to the ones used in the Ar
due to that natural Kr contains 11.5% of the magnetically active matrix simulation as shown in Fig. 1(a) and Table I.
83Kr isotope that contributes to the superhyperfine broadening As a result, the correlation time τcorr = 28.5 ns is some-what
admixing in addition Gaussian profile to the EPR lineshape. greater compared to Ar matrix. Solid Ne is a tighter matrix com-
On the basis of the hf component amplitude ratio12 and the pared to the solid Ar as seen from the lattice parameters of 4.46 Å
experimental line widths,19 an excellent EasySpin simulation is for Ne and 5.31 Å for Ar. As discussed above, a tighter matrix
presented in Fig. 1(b). The line widths of the simulated spectrum, enables smaller radical libration angles, and less averaged anisotro-
0.0275, 0.0234, 0.0247, 0.0313 mT, counting from the low- to the pies, yielding larger ΔA. The difference between the CH3 radical hf
high-field component, match extremely well with the experimen- A-tensor anisotropies in Ne and Ar would be more pronounced if
tal data:19 0.0301, 0.0218, 0.0241, 0.0334 mT. The EPR parameters it were not for the semi-quantum nature of solid Ne, characterized
used in the simulations for CH3 in Ar, Kr, Ne, and para-H2 are by fairly large zero-point displacement amplitude of the Ne matrix
listed in Table I. The isotropic A- and g-tensor components were atoms.
set equal to those of CH3 in solid gases from the previous study.13 A very small difference in the relative amplitudes of the CH3 hf
It was found that both the hf anisotropy, ΔA = A|| − A⊥, and the components was recorded for methyl radical trapped in solid
g-tensor anisotropy, Δg = g|| − g⊥, should be modified for best para-H2.12 In this case, even a small error in p
experimental
ffiffiffiffi amplitudes
agreement of the simulations with the experiment. may result in rather large uncertainty in 1= A as a function of the
projection of the total nuclear spin. This is evident from Fig. 1(d)
where the experimental points deviate more from the theoretical
fitting curve compared to the noble gas curves in Figs. 1(a)−1(c).
TABLE I. EPR parameters for CH3 in solid gases used in the spectral The deviation taken in the percentage of the overall spread of the
simulations in Fig. 1. Isotropic hf coupling and g-tensor parameters are hf-line amplitudes is addressed here.
taken from Ref. 13. Notice the trend of decreasing rotational correlation The EPR transitions recorded in the H2–CH3 experiment12
time with increasing atomic radius of the host. were rather broad, with a peak-to-peak width of the third, mF = −1/2,
component 0.022 mT compared to approximately 0.01 mT
Aiso, ΔA, τcorr, expected for radicals isolated in magnetically silent matrices.
Matrix mT g
iso mT Δg ΔH, mT ns Indeed, the width of the third EPR hf component in solid Ar that
Ne –2.333 2.002526 0.108 –1.844 10–4 0.0084(L) 28.5 has not magnetic isotopes is 0.014 mT. The small, extra broadening
Ar –2.313 2.002322 0.098 –1.73 10–4 0.0086(L) 24 in para-H2 is probably due to admixture of a small amount of
Kr –2.300 2.001655 0.138 1.1310–4 0.0045(L) 17 ortho-H2 molecules generated during the para-hydrogen radiofre-
0.0156(G) quency discharge before sample condensation. The profiles of the
H2 –2.324 2.002516 0.1 –0.610–4 0.0045(L) 2.3 simulated transitions were obtained as mixtures of Gaussian and
0.0196(G) Lorentzian components, applying 0.0045 mT Lorentzian width
taken from the CH3–Kr experiment. On the basis of this value, the

Low Temp. Phys. 45, 000000 (2019); doi: 10.1063/1.5093524 45, 000000-431
Published under license by AIP Publishing.
 Low Temperature ARTICLE scitation.org/journal/ltp
Physics

experimental width of the Gaussian contribution to the lineshape scales less than 10−10 s are required to fully average the magnetic
was calculated as 0.0196 mT. The closeness between the amplitudes parameters of the probe, e.g., the nitroxide. Libration motion is also
of the hf components for CH3 in the para-H2 lineshape made it efficient averaging mechanism at the EPR time scale. The term
difficult to select spectral parameters for best fitting. On the basis libration is commonly used for harmonic angular oscillations of
of results for the noble gas matrices, the estimated hf tensor anisot- molecules in crystals with frequencies of 1011 to 1012 s–1 and ampli-
ropy, ΔA, were of the order of 0.1 mT compared to 0.098 mT in tudes of approximately 2–3 deg. A similar type of motion in glasses
Ar, 0.138 mT in Kr, and 0.108 mT in Ne. has been evidenced by high-frequency EPR, magnetization transfer,
As it was verified in the present study, the hf tensor anisotropy and spin-echo experiments.23,25 However, low-temperature tor-
ΔA was not that sensitive to the matrix particles surrounding the sional oscillations greater than 8 deg were detected in the present
radical species as was the g-tensor anisotropy. The nearest coordi- work for tunneling methyl isolated in solid CO2, vide infra. This
nation distance of methyl to the matrix particles in solid H2 and Ar motion seems to correlate with unusually wide “hopping preces-
was nearly the same: 3.76 Å. The hf anisotropy ΔA in solid H2 is sion” angles of the linear CO2 molecules discovered recently by
expected to be somewhat smaller than in Ar due to the quantum Krainyukova and Kuchta5 in frozen samples of pure CO2.
nature of solid hydrogen while on the contrary, the axial hcp Due to libration, the anisotropies of methyl spectra were
(hexagonal close packed) crystal structure of H2 favors increased ΔA. partially averaged as shown earlier,20 gradually decreasing with
The best fit, Fig. 1(d), was obtained with Δg = −0.6⋅10–4 and decreasing eccentricity of the matrix molecules in the order: CO2,
τcorr = 2.3 ns. The g-tensor anisotropy turned out to be the smallest N2O, CO, N2. Our analysis showed that the greatest uncertainty in
among the matrices analyzed. This is because of more efficient obtaining rotational correlation times was associated with the
averaging of the anisotropy by the significantly increased libration approximation of the obtained polynomial coefficients in Eq. (3).
motion of the radical in the H2 quantum crystal. The correlation It turned out that the relative errors were least for a, while b and c
time is far too shorter than for other matrices, since the trapped showed somewhat larger error than a. As a result, the EasySpin
radical is much freer to rotate. simulations yielded cautious estimate of τcorr error of about 4 ns rel-
ative to the shifts in b and c coefficients.
Figure 3 shows the rotational correlation time plotted against
2.2. Motional averaging mechanisms of EPR the squared Pauli repulsion energy Ep2 . The figure establishes a
anisotropy linear correlation between rotation hindering and repulsion of the
Methyl radical reorientational motions about the C3- and trapped CH3 radical by the matrix particles. It is suggested in the
C2-symmetry axes average partially the spin-Hamiltonian anisot- present study, that the averaged spectrum anisotropy is not only a
ropy depending mostly on the strength of the matrix-impurity function of τcorr and the radical–matrix interaction but depends
Pauli repulsion.12 The residual anisotropy may be observed in also on the lattice symmetry of the matrix. The influence of the
different amplitudes and widths of any of the hyperfine compo- matrix symmetry on the anisotropy of the radical EPR parameters
nents, symmetric with respect to the base line, or as anisotropic is readily seen from the experimental and theoretical studies of
splitting. The former type of the anisotropy is common of CH3 in trapped H atoms.
molecular hydrogen H2 and solid noble gases Ne, Ar, Kr,12,27 while
the later was observed in solid gases consisting of the linear CO,
N2O, CO2 molecules.13,14,20 The rotational averaging is effective
even at liquid helium temperatures due exclusively to tunneling.
The present study is aimed at correlating tunneling rates to barriers
of rotation for CH3 in van der Waals solids, applying theoretical
approaches describing the experimentally observed residual EPR
anisotropy of trapped radicals.
Lack of direct experimental evidence of anisotropy even at
liquid He temperatures is the case for the alpha-protons of free
methyl, as only the high resolution 1:1:1:1 ground rotational EPR
quartet is visible.11 Pure quantum mechanical inertia rotation treat-
ment required that methyl rotor stops at this temperature with only
the above-mentioned spin-rotation state possible.11 Therefore, all
changes of the temperature-dependent hf line splitting were consid-
ered to be due to dynamics. Theoretical values of static magnetic
parameters of methyl in vacuum coming from first principle
quantum chemistry computations can be of importance for this
purpose.13 A great help for the disentanglement of the parameters
was offered by the simulation of the high-temperature limit spectra
that do not exhibit complications due to motional quantum effects.
The effect of molecular reorientation on the EPR lineshape is FIG. 3. Correlation time of the perpendicular rotation of CH3 radical in various
matrices plotted against the squared Pauli repulsion energy Ep2, where ε0 ¼
commonly involved in quantitative methods of molecular mobility 27:212 eV, is the Hartree atomic unit of energy.
data extraction. In low viscosity liquids, angular velocities of time

Low Temp. Phys. 45, 000000 (2019); doi: 10.1063/1.5093524 45, 000000-432
Published under license by AIP Publishing.
 Low Temperature ARTICLE scitation.org/journal/ltp
Physics

Free atomic hydrogen is characterized by spherically symmet-

rical electron wave function and isotropic hf constants. When
trapped in solids, H atom reveals wave function of reduced symme-
try adapting to the electrostatic potential of the cage formed by the
matrix first couple of coordination spheres. As a consequence,
excluding cubic lattice, the hf interaction develops appreciable
anisotropy.28–31 The excellent linear fit in Fig. 3 may be accidental.
Further experimental work with Kr matrices with unstrained
crystal lattice, condensed, e.g., at temperatures well above 4.2 K, is
required for the verification of the linearity.

3. ROTARY TUNNELING VS. HINDERING BARRIER OF

METHYL RADICAL
Even strongly hindered methyl groups can perform rotary
motion at very low temperature in the sense that quantum FIG. 4. Correlation diagram showing roughly the connection between the three
mechanics allows penetration of the torsional barrier. A periodic different types of rotary/torsional motions. Notice that the three vertical axes
C3 potential with periodic boundary conditions will be further cannot be put in a universal scale since the transition frequencies is a multivari-
considered hindering coherent leakage of methyl to the adjacent ate quantity. The ground rotation level has been lifted by the zero-point vibration
energy E0 = 3(BV3)1/2/2. The first order correction of the torsional level amounts
minima with “distance” 120 deg. The rate of this rotary motion of to C0, see Appendix. The tunneling frequency of the ground vibrational level is
methyl can be faster than the proton hfi (hyperfine interaction) designated by 3J0. Going from the left to the right, the second diagram corre-
measured in MHz. The situation can be the reverse for systems sponds approximately to the librating radical, while the third diagram to the right
with high barriers. The deuteron rotor, e.g., is normally much is for the radical performing rotation around the C3 axis by tunneling through the
slower, vide infra, due to the “dramatic” increase of the inertia by barrier with frequency 3J0. In the end of this section, the description of this
factor two. motion physics is provided.
The point here is that one has to use a quantum description
of the motion in order to reproduce the experimental effects on the
EPR lineshape, at least for the lowest temperatures close to 5 K. E.g., the overlap of the lowest (ground) torsional level with
We can use, e.g., approximate harmonic oscillator states localized quantum number ν = 0 is given by
at the minima of the potential energy wells for the description of
( rffiffiffiffiffiffi)
this motion, see Appendix. Certainly, Hermite polynomials, or any π2 V3
other simple choice like Gaussians,32 cannot be accurate eigenfunc- S0 ; hΨ(k) (kþ1)
0 (w)jΨ0 (w)i ¼ exp  : (5)
6 B
tions of a Hamiltonian with an actually periodic potential but they
can be used as (zeroth order) basis functions for the approximate
solution of the problem. This equation is plotted in Fig. 5 and indicates that the overlap is
Briefly, the approximate localized functions of the smallest C3 not really significant in the present case.
group are giving initially an equal number of degenerate basis-
functions Ψ(k)v (w) ; Ψv (w  ϑk ) with location k = 0, ±1, seen in
(k)

the correlation diagram, Fig. 4. The angles ϑk = kp are given in

terms of the period of the potential energy p = 2π/3. The index
ν = 0, 1, 2, … of the above functions stands for the vibration level.
The eigenvalues of these functions are given by
 rffiffiffiffiffiffi
1 3 V3
Ev ¼ 2β B v þ , β ¼ , (4)
2 2 B
pffiffiffiffiffiffiffiffiffiffiffi
where B is the rotational constant and the variable β ¼ (3=2) V3 =B
is expressed in terms of the just mentioned constant Β and the
barrier V3, as seen in Appendix A. The 3-fold degeneracy of the
above-localized vibrator eigenfunctions can be lifted by a periodic
potential if the barrier between adjacent wells is not infinitely
high. The problem can be solved by both degenerate perturbation
or by the variational method.33 The variational principle is for- FIG. 5. The overlap S0 of the adjacent vibrator eigenfunctions for the ground tor-
mally more appropriate because the basis functions of the local- sional level of methyl according to Eq. (5). Notice that the value of this integral is
ized torsional oscillators cannot be orthogonal as they overlap. As very small already for potential bariers V3 > 100 K and becomes virtually insignifi-
cant after 250 K. Those potentials are in compatible order of magnitude for methyl
it will be shown later, the overlap, however, is small for all practi- rotor systems studied in the present work (1.000 K = 2.0837⋅104 MHz).
cal purposes.

Low Temp. Phys. 45, 000000 (2019); doi: 10.1063/1.5093524 45, 000000-433
Published under license by AIP Publishing.
 Low Temperature ARTICLE scitation.org/journal/ltp
Physics

In spite of that, the solution of the tunneling frequency vs. Finally, the mechanism of the tunneling rotation of the radical can
potential problem obtained by this way will not be as accurate as a be described as a quantum effect by the time-dependent Schrödinger
numerical solution model starting from free rotor in presence of a equation. The diagonalizing transformation of the localized vibrator
periodic potential. In practice, it was not easy to make the solutions eigenfunctions leading to the torsional oscillator eigenfunctions can
of the perturbative kind accurate enough, see Fig. 6(a). However, be inverted and the former set can also be used in the time depen-
the motivation of seeking such a solution, in addition to relatively dence of the system. In that case, if the initial state of the system is in
easily obtainable numerical ones, is rather the transparency and one of the potential wells, it will be found in any of the two adjacent
the usefulness in the physical interpretation of several effects. wells at a time equal to the inverse energy difference of the E–A
Symmetry considerations play a very important role here. states, signifying the quantum tunneling of the system over the
Further, one should remember that the analytical expressions potential barrier.
in the present work are approximate. The assumptions are: i) con-
sideration of parabolic zero order potential; ii) neglect of the
mixing with higher vibrational levels. 3.1. Computation of the tunneling frequency
The parabolic potential was obtained by Taylor–Maclaurin One of the aims of this work is to derive analytical expressions
series expansion of the periodic C3 potential for the tunneling frequency 3J0 vs. the potential barrier V3. Hence,
a combination of the perturbation and the variational methods
V3 were tested. The value of the overlap integral Sv ¼ hΨ(k)v jΨv
(kþ1)
i of
V(w) ¼ [1  cos 3w]: (6) any two adjacent localized torsional oscillator eigenfunctions was
2
originally omitted in the computation of the perturbation treatment
because it is very small for the case here V3 >> B. This is at least
verified for the lowest vibrational levels as shown in Fig. 5, which is
a plot of the overlap vs. barrier shown in Eq. (5).
The inclusion of the overlap in a more accurate analytical
expression for the torsional splitting of the ground level according
to the variational theory has also been investigated. When this
effect was included it did not help practically, because on the con-
trary, for low-potential barrier where the overlap is important the
analytical expression obtained by perturbation gave better agree-
ment with the numerical result. For high-potential barrier where
the disagreement worsens for the perturbation method, the overlap
is very small, and the variation method yielded the identical result
as the perturbation.
Also, ENDOR has been used to determine tunneling fre-
quency. The theoretical treatment of the shift of ENDOR transi-
tions due to tunneling is usually obtained by using the second
order perturbation theory of an effective spin-Hamiltonian.34,35
The potential barrier is then deduced by numerical simulations
involving a series of barriers and computing the energy difference
between the two lowest levels.34 Hence, we tried to obtain the most
natural analytic expression of the tunneling frequency vs. the
potential barrier described further in the text.
Figure 6 shows the tunneling frequency of the parallel rotation
of both protonated and deuteriated methyl radicals vs. potential
barrier. The numerical method is based on diagonalizing the pure
rotational Hamiltonian including a hindering periodic C3 potential
in the basis set of a two-dimensional free rotor. Similarly, the anal-
ysis of the tunneling rotation about the in-plane axes is based on a
C2 symmetry potential with periodic boundary conditions.
The analytical computation of the C3 case is based on the
degenerate perturbation treatment of localized torsional Hermite
FIG. 6. (a) CH3 and CD3 tunneling frequencies about the C3 axis for different
values of the barrier V3 keeping the value of the kinetic constants B = 6.752 and polynomial oscillator eigenfunctions. Both the barriers of the pro-
3.376 K for protonated and deuteriated methyl, respectively (1 K = 20.837 GHz). tonated and the deuteriated methyl radicals were computed and
The upper limit of the tunneling frequency for V3 = 0 is formally equal to the compared. One good reason for studying also the deuteriated CD3
rotational constant B. (b) Plot of the numerically obtained tunneling frequency values is that often the EPR spectra of deuteriated methyl are
vs. potential barrier of methyl and deuteriated methyl about any in-plane C2 misinterpreted in literature. Look for the experimental EPR data
axes of the radical. The computational method was parallel to the approach for
the C3 case. The details of the method are described in Appendix C.
along with the corresponding accurate theoretical predictions
shown in Ref. 11.

Low Temp. Phys. 45, 000000 (2019); doi: 10.1063/1.5093524 45, 000000-434
Published under license by AIP Publishing.
 Low Temperature ARTICLE scitation.org/journal/ltp
Physics

In Fig. 6, the method used in the computations, either numer- were used to span the Hamiltonian matrix in the two-dimensional
ical, or deriving analytic expressions using the perturbation theory rotational space of a full cycle for a rotor influenced by the simplified
on assumed localized oscillators is indicated. The Hermite polyno- hindering potential seen in Eq. (6). The matrix representation of this
mials are eigenfunctions of these oscillators inside each potential Hamiltonian was physically blocked by symmetry in three different
well of the two- or three-fold symmetry axis of rotation. The diago- sub-matrices, an A-symmetry block accompanied by an Ea and an
nalization of the following matrix for the rotational Hamiltonian in Eb symmetry blocks. The computational details of the matrix ele-
presence of the C3 potential V(w) in Eq. (6) will result in the cor- ments and the computations of the tunneling frequency vs. barrier
rected, albeit approximate, energies for the torsional oscillator V3 are shown in Appendix B and visualized in Fig. 6.
0 1 This way of working corresponds to separating the three-
Ψ(1)v Ψ(0)v Ψ(þ1)
v dimensional (3D) quantum rotator considered by the Popov and
B Ψ(1) Dv Jv Jv C coworkers24 into either a C3 or a C2 methyl-axis rotation under
Mv ¼ B
@ Ψ(0)
v C:
A (7)
v Jv Dv Jv different hindering conditions. They studied in detail how different
Ψ v
(þ1)
Jv Jv Dv orientations of the radical rotation axis with respect to a cubic
matrix cage determine the strength and symmetry of the hindering
The diagonal matrix elements Dν in the isolated oscillator basis barrier. In practice, matrix isolated methyls are embedded in
set contain the regular quantum vibrator energy Eν and the first well-defined host cavities restricting the direction of rotary motion
order correction Cν due to the potential V(w), see Appendix. not allowing free 3D motion. The restriction of the angular
The off-diagonals matrix elements Jν between adjacent oscillators momentum projection with respect to a laboratory-defined frame
are also shown there. Notice that their nature, the number, and reduces significantly the rotational degeneracy even in cases where
the presence even at the outer off-diagonals positions of matrix the rotation appears practically free.17,20 The numerical method of
Mν in Eq. (7), incorporate the physical meaning of periodic boun- the present work can easily be extended to involve even more
dary conditions, necessary for the present case. sophisticated potentials, exploiting the work of Popov and cowork-
Diagonalizing the above matrix, the eigenvalues EA = Dν + 2Jν ers,24 provided that the matrix elements of the relevant potential
and ΕΕ± = Dν − Jν of the system, corresponding to the totally sym- can be computed in the given exponential basis set.
metric A, and the doubly degenerate E-symmetric torsional level, The explicit analytical expressions of the present work are not
respectively, were obtained. The second order tunneling frequency valid for very small V3 < 50 K because of the method of approxima-
fν,tun amounts to the difference tion. However, for very low barriers of the order of 20–25 K and
lower, it is expected that the “tunneling” frequency will be close to
fv ¼ EE + ,  EA ¼ Dv  Jv ,  (Dv þ 2Jv ) ¼ 3Jv (8) the rotational constant B as “free” rotation is approached and
which is shown in the next equation explicitly in terms of the formal equality to B will be valid even at liquid He temperatures.
barrier V3 and the rotational frequency B In spite the qualitative agreement of the above expression in
Eq. (9) with the numerical computation, the two series of values of
" rffiffiffiffiffiffi!#
3V3 π2 V3 the tunneling frequency follow each other rather well in the whole
(2) (2)
f0,tun ¼ 3J0 ¼ exp  interval of the quotient V3/B. The trend of the numerical computa-
2 6 B
tions for the two different rotors, proton, and deuteron, are also

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  rffiffiffiffiffiffi 
3π 2
9 B 3 B followed in the whole interval of the independent variable V3/B,
 1  þ exp  : (9) see Fig. 6.
2 2 V3 2 V3
However, the analytical values underestimate the numerical
This expression for the tunneling splitting of the ground level is tunneling frequency more and more for larger values of the quotient
usually designated by 3J0.30 In the present work, the tunneling V3/B. One should expect the best agreement for large values of V3
frequency is obtained by including the contribution of the full relative to B where the overlap is the least. In reality, this statement
second-order effects concerning the periodic potential, as shown holds for the absolute values of the two different computational
in Appendix. results, while progressively along with increasing V3, the analytical
There are several treatments in the literature with varying value of the tunneling frequency worsens in percent units.
terminology where the symbol J0, or similar, has been used identi- A formally more appropriate method than perturbation
cally as the hΨ(k)
v (w)jV(w)jΨv
(kþ1)
(w)i matrix element and/or as an should be the variational method, irrespective of the kind of the
empirical constant for the tunneling frequency in a pure spin- used trial functions Ψ. The reason is that the basis functions that
Hamiltonian. To distinguish with the present full second-order were used for the perturbation treatment were not orthonormal,
degenerate-perturbation treatment, we will use the symbol J0 for because the eigenfunctions of the well oscillators used are partially
0
the off-diagonal matrix element hΨ(k) (kþ1)
v (w)jH (w)jΨv (w)i for the overlapping for the potential barriers of interest. According to the
perturbation part of the rotational Hamiltonian HR = H0 + H0 , seen variational
Ð method,
Ð minimization of the integral quotient
in the Appendix, Ref. 36. Ψ*HR Ψdw= Ψ*Ψdw, should be Ð performed. However, consider-
According to another interpretation, the tunneling frequency is ation of the missing overlap Ψ(k)* Ψ(k+1) dw, of the localized
the energy difference of the two lowest levels of the above-hindered Hermite-polynomial basis used in the computations of the per-
rotation Hamiltonian. The relevant numerical computations were turbation did not improve the result. At least, according to the
performed in the more convenient basis set of the normalized, Bloch variational method, only an upper estimate of the energy can
type, exponentials { exp (  inw)=2π:n [ Z + , w [ [  π, π]}. They be achieved.

Low Temp. Phys. 45, 000000 (2019); doi: 10.1063/1.5093524 45, 000000-435
Published under license by AIP Publishing.
 Low Temperature ARTICLE scitation.org/journal/ltp
Physics

As seen also in the plot of Fig. 5, the values of the over-lap

integral are negligible for the barriers of interest. A test of a more
careful consideration according to the above lines did not improve
the result, thus justifying the present investigation. It remains
then to give an explanation and a suggestion for improving the
analytical result.

4. PARALLEL AND PERPENDICULAR METHYL RADICAL

TUNNELING FREQUENCIES
The periodic C2 and C3 hindering potentials used in the com-
putations of the present work is an approximation that simply indi-
cates tunneling rotation about the in-plane axes or the higher
symmetry axis of methyl radical, respectively. The obtained results
on the tunneling-frequency vs. rotation-barrier are similar but the
magnitude of the barrier and the periodicity of the potential deter-
mined the final details. An improved approach for more realistic
potentials was presented by Popov, Kiljunen et al.,8,24 quantifying
appropriate structural/geometrical parameters of the system, poten-
tially useful for numerical simulations. Using the group theory,
they incorporated the particular cubic cage geometry and the Oh
symmetry of the host in combination to the intrinsic C3 methyl
symmetry into the final form of the hindering potential, where in
addition radial conditions were involved.
Figure 7 and Table II show results based on the analytical C2
treatment. The correlation times obtained from the EPR simulation FIG. 7. Computed tunneling frequency vs. potential in K for C2 rotation (blue
of CH3 in matrices of the spherical particles Ar, Kr, Ne and p-H2, dash −), in the range of six different, experimentally determined, tunneling fre-
quencies. The values in Table II were obtained by fitting the experimental tun-
presented in Figs. 1(a)–1(d), were used to derive the potential barri- neling frequencies the range 250 K < V3 < 850 K to the exponential function ftun
ers. Figure 7 also shows an approximate exponential graphical fitting = 28715 exp(−0.01V2) MHz. The agreement with the numerically obtained
of the numerically obtained potential barriers from the tunnelling values for C2 symmetry tunneling with the straight line featuring the logarithm of
frequencies. Diagonalization of the hindered rotation Hamiltonian the latter exponential relation was remarkable.
was used in both the above C2 rotation case as well as for the follow-
ing C3 case.
Figure 8 shows the computed barriers corresponding to obtained tumbling frequencies, 2.0⋅107 and 1.3⋅107 s–1 for CH3
observed tunneling frequencies obtained from the experimental and CD3, respectively. Although rapid, these tumbling frequencies
spectra simulations for linear and spherical host molecules discussed still indicate a considerable hindrance to methyl rotation by the
in the present study and also in Refs. 2 and 4, where estimations of silica gel surface, as free rotational frequencies are usually by
possible C3 symmetry barrier values are found for a comparison. three orders of magnitude (103) faster. The hindering potential
The matter of the fact is that the frequency of 1.4 MHz is found for barriers determined by the presently developed barrier/tunneling
rotation about the in-plane C2 axes and concerns methyl radicals in relations indicate in particular barriers of ca. 678 K for the pro-
matrices of linear molecules: CO, N2O, CO2. In this case, a barrier of tonated and 326.8 K for the deuteriated methyl radical.
1012 K was obtained as shown in Table II. However also a C3 tunnel- It seems that the eccentricity of the matrix molecules contrib-
ing of this type of systems gives the same order of magnitude barri- utes to both the form and the height of the barrier and should be
ers, ca. 946.2 K, as obtained from the graphical representation of
Fig. 8. On the other hand, tunneling frequencies ca. 0.1 GHz are
obtained for matrices of spherically symmetrical particles: para-H2, TABLE II. The obtained correlation times considered as invert tunneling
Ne, Ar, Kr. Here, the corresponding barrier for C3 tunneling is frequencies of the CH3 rotor isolated in four solid gases in the first section
approximately 534.9 K as seen in the graphical representation of of this work along with numerically computed potential barriers V3.
Fig. 8, which is in good agreement with 585.8 K for the C2 tunneling
shown in Table II. System τcorr, ns ftun, MHz V3, K
The EPR spectra of both CH3 and CD3 adsorbed on the Linear host CO, N2O, CO2–CH3 710 1.4 1012
silica gel surface at 77 K were studied by Gardner and Casey.37 Ne–CH3 28.5 35.1 690.5
The dependence of the line width on the nuclear spin quantum Ar–CH3 24 41.7 673.3
number was interpreted in terms of the tumbling of the radicals Kr–CH3 17 58.8 638.9
on the surface and the values of the experimental correlation Spherical host para-H2, Ne, Ar, Kr ∼10 ∼100 585.8
times/tumbling frequencies were determined as a function of the pH2–CH3 2.3 435 438.8
hyperfine A- and g-tensors anisotropy. Gardner and Casey

Low Temp. Phys. 45, 000000 (2019); doi: 10.1063/1.5093524 45, 000000-436
Published under license by AIP Publishing.
 Low Temperature ARTICLE scitation.org/journal/ltp
Physics

matrices with larger eccentricity, such as N2O and CO2, show well
discerned weak-line multiplet.

5. CONCLUSION
We first obtained rotation correlation times (tunneling fre-
quencies) of the CH3 radical in matrices of spherically symmetrical
host particles held at liquid helium temperatures. The tunneling
rates are correlated to the radical–matrix coupling leading to rota-
tion hindering, mainly governed by the Pauli exclusion forces
between trapped radical and matrix molecules.
Matrix shifts of spectrum anisotropies in trapped radicals
testified that matrix effects are more important on Δg than on ΔA.
Quench-condensed Kr crystallizes most probably into axially sym-
FIG. 8. Graphical representation of the numerically computed barriers vs. tun- metric hcp local structure, surrounding the trapped CH3 radical.
neling frequency for rotational frequency parameter of planar methyl B = 6.752
The theoretical treatment of this study explains how the fast
K. Left for tunnel frequency 0.1 GHz = 100 MHz barrier V3 = 534.9 K, middle for
frequency 20 MHz barrier V3 = 678 K, and right, for frequency 1.4 MHz barrier orientational motion of methyl at low temperatures averages to
V3 = 946.2 K. different degrees the anisotropy of certain magnetic parameters due
to rapid rotary tunneling about its C3 and C2 symmetry axes allow-
ing estimation of the potential barriers to the rotation.
A clearly higher barrier of methyl rotation for the radical iso-
utilized in the determination of the potential barrier characteristics. lated in matrices of linear vs. spherical matrix molecules was
It is certainly correlated to the libration motion or the hoppling pre- obtained by evaluating the barrier of experimentally obtained tun-
cession of linear molecules in the sense put forward by Krainyukova neling frequencies. In particular, the linear CO, CO2, N2O molecu-
and Kuchta,5 which does not occur in case of matrices of spherically lar matrices seem to hinder tunneling methyl more efficiently than
symmetrical particles. During the libration or hopping precession, matrices of spherical molecular particles such as Ar, Kr, Ne and
a matrix molecule sweeps a cone which thus is not available to the p-H2. The barriers that the above linear molecular matrices oppose
methyl radicals. Evidently, the larger the molecule eccentricity, are slightly over 1000 K for both C2 and C3 rotational tunneling,
the larger is the cone volume, i.e., the barrier width and height. which is almost double the barriers opposed to methyl tunneling
Although this picture is not directly suitable to apply in a mathe- by the above-mentioned spherical ones.
matical model for the potential energy, it may be considered as a
direction to follow in understanding the role of the eccentricity.
Namely, the occurrence of such a complicated EPR spectrum ACKNOWLEDGMENTS
of the CH3 radical in solid N2O and CO2 matrices, in contrast to N.P.B. is thankful to professor Nikolaos Kyratzis for assistance
solid rare gases and in N2 and CO matrix isolation, depends on the in certain numerical methods necessary for the best approach for
relation between the structure of the hosted radical and the lattice the presentation of some parts in this work. Yu.A.D. and I.A.Z.
molecules.4 An appropriate quantity reflecting our qualitative con- acknowledge support of the Russian Foundation for Basic Research
siderations is the host molecule eccentricity ε taken simply as the (RFBR), research project 16-02-00127a, for the experimental part
ratio of the maximum of the internuclear distance d to the lattice of the study.
parameter a. The results are summarized in Table III together with
the relevant parameters supporting the above presumption. Indeed,
no weak orthorhombic satellites were observed at temperatures APPENDIX A. TORSIONAL OSCILLATOR EQUATIONS
above 8 K in CO and N2 matrices2,14,20 which have relatively small
The zero order eigenfunctions of the torsional oscillator for a
eccentricity, neither such spectra were reported for matrices of
three-fold periodic potential, with period p = 2π/3, and periodic
spherical symmetrical host particles. On the other hand, the
boundary conditions, can be approximated by the following local-
ized harmonic oscillator eigenfunctions:
xk =2
2
TABLE III. Estimation of the molecule eccentricity for various cryocrystals v (w) ¼ Nv Hv (xk )e
Ψ(k) : (A:1)
at 20 K. The parameters d and a are available from Ref. 1.
Τhe Hermite polynomials Hν have the scaled angular coordinate
Internuclear distance Lattice parameter Eccentricity argument xk, given by
Matrix d, Å a, Å ε = d/a pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
xk ¼ β(w  ϑk ) (A:2)
N2 1.098 5.658 0.193
CO 1.128 5.652 0.200 with a specification for the kth well position by the angle ϑk = kp
N2O 2.312 5.641 0.410 numbering the potential energy minima at the wells by k = 0, ±1,
CO2 2.320 5.554 0.418 where p = 2π/3 is the period. The k-independent normalization
factors Nν of the above eigenfunctions are given in the following

Low Temp. Phys. 45, 000000 (2019); doi: 10.1063/1.5093524 45, 000000-437
Published under license by AIP Publishing.
 Low Temperature ARTICLE scitation.org/journal/ltp
Physics

equation: APPENDIX B. DERIVATION OF MATRIX ELEMENTS

 pffiffiffi 1=2 The following matrix elements were used in the treatment
β mentioned in the above Appendix A concerning the ground har-
Nv ¼ v pffiffiffi : (A:3)
2 πv! monic vibrational level. They are the diagonal (B.1) and the position
overlap (B.2) of the square of the angular relocation variable w.
The constant β is the square of the angle-scaling factor of the argu- The expression for the variance of the angle w of the level ν
ment xk given by for the localized vibrator given by
rffiffiffiffiffiffiffiffiffi pffiffiffi
3 V3 (2v þ 1) B
β¼ , (A:4) hΨv (w)jw jΨv (w)i ¼
(k) 2 (k)
p ffiffiffiffiffi
ffi : (B:1)
2 B 3 V3

where V3 is the potential barrier and Next, the off-diagonal matrix elements of the square of the angular
relocation variable w is given
B ¼ (h=2π)2 =2I (A:5) sffiffiffiffiffiffiffiffiffiffiffi
pffiffiffi ffi rffiffiffiffiffiffi!" rffiffiffiffiffiffi#
B π2 V3 π2 V3
hΨv (w)jw jΨv (w)i ¼
(k) 2 (kþ1)
pffiffiffiffiffiffi exp  1þ :
is the rotation constant. This parameter represents the rotational 3 V3 6 B 3 B
frequency of planar methyl and has approximately the value 6.752 K
for the protonated and half that value, 3.376 K, for deuteriated (B:2)
methyl radical species. The denominator of Eq. (A.5) gives the par- The above two matrix elements were further computed for the full
allel component of the methyl rotor moment of inertia I, corre- hindering potential V(w):
sponding to rotation about the highest symmetry C3 axis of a   rffiffiffiffiffiffi
planar methyl. V3 3 B
C0 ¼ hΨ(k) (w)jV(w)jΨ(k)
(w)i ¼ 1  exp  : (B:3)
The rotational constants according to Prager and Heidemann38 v v
2 2 V3
are B(CH3) = B(CH4) = 0.655 meV = 7.601 K (1 eV = 1.1605⋅104 K), a
value corresponding to the nonplanar –CH3 fragment, that agrees This term is the first-order potential energy correction to the energy
with the one adopted in our previous work.11 Popov et al.24 and E0 in Eq. (A.7) of the ground vibrator eigenfunction, obtained as a
Kiljunen et al.8 consider instead as rotational constant the double B particular case of the following general ν level vibrator correction:
value, equal to the inverse moment of inertia. Cv ¼ hΨ(k) (k)
v (w)jV(w)jΨv (w)i
The above relations are based in the Taylor expansion of the "  rffiffiffiffiffiffi v  2l  #
potential V(w) (6) about its three minima. The resulting approxi- V3 3 B X (  1)l 3 v
¼ 1  exp  pffiffiffiffiffi : (B:4)
mate harmonic potential V0 (w) = 9V3w2/4 is seen in the following 2 2 V3 l¼0 l! 2β l
differential operator, used as the unperturbed part of the torsional
Hamiltonian at each well: On the other hand, except for the ground level, the general expres-
sions of the higher level off-diagonal matrix elements of the potential
1 are particularly difficult to obtain. For the ground level we have
H0 ¼ B@ 2 =@w2 þ (9V3 =2)w2 : (A:6)
2
J0 ¼ hΨ(k)
v (w)jV(w)jΨv
(kþ1)
(w)i
The following energies were obtained for the localized torsional rffiffiffiffiffiffi!  rffiffiffiffiffiffi
levels: V3 π2
V3 3 B
¼ exp  1 þ exp  : (B:5)
2 6 B 2 V3
Ev ¼ 2β B(v þ 1=2): (A:7)
The diagonal matrix elements for a general vibrational level ν were
Notice that this is a harmonic oscillator Hamiltonian that approxi- first computed by paper and pencil and were reduced for the ground
mates the periodic potential energy by a parabola about each level by putting ν = 0. The matrix elements were then verified by
minimum. However, because the periodicity is not present in the comparing to expressions obtained with “mathematica”.
approximate V0 (w), an additional task is to imply it at a later stage.
The perturbation applied to the above zero order Hamiltonian
was APPENDIX C. NUMERICAL TREATMENT
0 The matrix elements of the quantum rotation Hamiltonian
H ¼ V(w)  9V3 w =4: 2
(A:8)
under the influence of a hindering potential HR ¼ B@ 2 =
The potential energy V(w) is given in the above Eq. (6). The result @w2 þ V(w), in the Bloch type, imaginary exponential basis set
obtained after the summation of the terms in Eqs. (A.6) and (A.8) {N exp ( in w):n [ Z+ , w [ [ π, π]}, are given by combining
is the identical total rotational Hamiltonian HR: the orthogonality relation
ðπ
HR ¼ B@ 2 =@w2 þ V(w) (A:9)
(eim w )*ein w dw ¼ 2 π δm,n ) hmjni ¼ δm,n (C:1)
that comprises the full periodic potential. π

Low Temp. Phys. 45, 000000 (2019); doi: 10.1063/1.5093524 45, 000000-438
Published under license by AIP Publishing.
 Low Temperature ARTICLE scitation.org/journal/ltp
Physics

pffiffiffiffiffi
with normalization factor N ¼ 1= 2π and the following matrix Another advantage of the above numerical treatment is that
elements. the higher than the ground level tunneling frequencies can also
The matrix elements of the differential operator in the rotation easily be obtained, because all the higher level eigenvalues of the
energy part of HR is given by above Hamiltonian are available by the identical diagonalizing
procedure.
ðπ The Hamiltonian matrix is in fact of infinite dimensions as
@ 2 in w
hmj@ 2 =@w2 jni ¼ N 2 (eim w )* e dw ¼ n2 δm,n : (C:2) n → ∞ in the basis set {| n > : n ∈ Z}. Therefore, in the computa-
@w2
π tions the matrix dimensions had to be increased until convergence
of the eigenvalues was obtained. Except for the matrix dimension
The following are the off-diagonal, C3 symmetry-dependent matrix dependent convergence procedure, for high potential barriers the
elements tunneling frequency is finally computed as the difference of two
very small numbers leading in an all greater numerical uncertainty.
ðπ Both the above error sources were most serious for deuteron com-
hmj cos 3 wjni ¼ N 2 (eim w )* cos 3 wein w dw putations as observed in the erroneous flattening of the numerical
π tunneling data of deuteron at the high V3 range of Fig. 6(a).
1
¼ (δm,n3 þ δm,nþ3 ):
2 REFERENCES
1
The only difference for the C2 case is the modified form of the V. G. Manzhelii and Yu. A. Freiman, Physics of Cryocrystals (AIP, New York,
above integral as follows: 1997).
2
Yu. A. Dmitriev, V. D. Melnikov, K. G. Styrov, and M. A. Tumanova, Physica B
ðπ 449, 25 (2014).
3
hmj cos 2 wjni ¼ N 2 (eim w )* cos 2 wein w dw Yu. A. Dmitriev, V. D. Melnikov, K. G. Styrov, and N. P. Benetis, Physica B
458, 44 (2015).
π 4
Yu. A. Dmitriev, I. A. Zelenetckii, and N. P. Benetis, Physica B 537, 51 (2018).
1 5
N. V. Krainyukova and B. Kuchta, J. Low Temp. Phys. 187, 148 (2017).
¼ (δm,n2 þ δm,nþ2 ): 6
2 V. V. Danchuk, A. A. Solodovnik, and M. A. Strzhemechny, Fiz. Nizk. Temp.
33, 783 (2007. Low Temp. Phys. 33, 600 (2007)]
7
Some additional symmetry related details are discussed in the end Yu. A. Dmitriev, Fiz. Nizk. Temp. 34, 95 (2008. Low Temp. Phys. 34, 75 (2008)]
8
of this section. T. Kiljunen, E. Popov, H. Kunttu, and J. Eloranta, J. Chem. Phys. 130, 164504
In the above given exponential basis set, the Hamiltonian (2009).
9
matrix HR in Eq. (A.9) with the simple C3 potential V(w) in Eq. (6) B. H. Bransden and C. J. Joachain, Physics of Atoms and Molecules, Longman
is found to be block diagonal. In fact, three separate banded tridiag- Scientific and Technical, (New York, 1983). P. 455.
10
E. O. Stejskal and H. S. Gutowsky, J. Chem. Phys. 28, 388 (1958).
onal matrices, one of A and two of E symmetry, were obtained. 11
T. Yamada, K. Komaguchi, M. Shiotani, N. P. Benetis, and A. Sørnes, J. Phys.
By increasing the dimension of the matrix to at least 60 × 60 overall, Chem. A 103, 4823 (1999).
an acceptable convergence of the eigenvalues was obtained. The 12
Yu. A. Dmitriev and N. P. Benetis, J. Phys. Chem. A 114, 10732 (2010).
tunneling frequency of the ground torsional level is the difference 13
N. P. Benetis, Yu. Dmitriev, F. Mocci, and A. Laaksonen, J. Phys. Chem. A
between the average EE = (Ea + Eb)/2 of the minimum degenerate 119, 9385 (2015).
14
eigenvalues of the E block minus the minimum non-degenerate T. Kiljunen, E. Popov, H. Kunttu, and J. Eloranta, J. Phys. Chem. A 114, 4770
eigenvalue EA of the A block. The computation of the tunneling (2010).
15
frequency was repeated as a function of the potential barrier V3 Yu. A. Dmitriev, G. Buscarino, and N. P. Benetis, J. Phys. Chem. A 120, 6155
and the C3 rotational constant B = 6.752 K. For the deuteriated (2016).
16
S. N. Foner, E. L. Cochran, V. A. Bowers, and C. K. Jen, Phys. Rev. Lett. 1, 91
methyl, half of this value for the rotational constant was used.
(1958).
On the contrary, the in plane C2 rotation of methyl requires 17
H. M. McConnell, J. Chem. Phys. 29, 1422 (1958).
double these values as rotational constants. This is because the per- 18
Yu. A. Dmitriev, Physica B 352, 383 (2004).
pendicular moment of inertia of a symmetric top disk is half the 19
N. P. Benetis and Yu. A. Dmitriev, J. Phys. Condens. Matter 21, 103201
parallel value. (2009).
20
Further, the potential energy function in Eq. (13) is different Yu. A. Dmitriev, V. D. Melnikov, K. G. Styrov, and M. A. Tumanova, Physica
in the simpler C2 tunneling case compared to C3. The Hamiltonian B 440, 104 (2014).
21
matrix in the C2 case is blocked in two banded tridiagonal matrices, EASYSPIN, free Matlab software written by S. Stoll, http://www.easyspin.org/.
instead of three, one corresponding to the totally symmetric A and Accessed August 26, 2016.
22
the other to the anti-symmetric B irreducible representations of the F. J. Adrian, J. Chem. Phys. 32, 972 (1960).
23
E. P. Kirilina, S. A. Dzuba, A. G. Maryasov, and Yu. D. Tsvetkov, Appl. Magn.
Abelian and cyclic C2 group.
Reson. 21, 203 (2001).
Notice that the above numerical method is flexible in another 24
E. Popov, T. Kiljunen, H. Kunttu, and J. Eloranta, J. Chem. Phys. 126, 134504
way, allowing the possibility of involving almost arbitrarily realistic (2007).
potentials. However, in some of these more complex systems the 25
A. K. Vorobiev and N. A. Chernova, “Simulation of rigid-limit slow-motion
simplifying block structure of the Hamiltonian matrix as for the EPR spectra for extraction of quantitative dynamic and orientational information,”
above simple C3 and C2 potentials may not be possible. in Intech, edited by A. I. Kokorin (2012), Chap. 3., 10.5772/74052.

Low Temp. Phys. 45, 000000 (2019); doi: 10.1063/1.5093524 45, 000000-439
Published under license by AIP Publishing.
 Low Temperature ARTICLE scitation.org/journal/ltp
Physics

26 33
Yu. A. Dmitriev, Fiz. Nizk. Temp. 33, 661 (2007) [Low Temp. Phys. 33, 493 I. N. Levine, Quantum Chemistry, 2nd ed. (Allyn and Bacon, Inc. Boston
(2007)]. 1974), p. 158.
27 34
E. Y. Misochko, V. A. Benderskii, A. U. Goldschleger, A. V. Akimov, H. C. Box, E. E. Budzinski, and W. R. Potter, J. Chem. Phys. 61, 1136
A. V. Benderskii, and C. A. Wight, J. Chem. Phys. 106, 3146 (1997). (1974).
28 35
K.-P. Dinse, Phys. Chem. Chem. Phys. 4, 5442 (2002). F. Bonon, M. Brustolon, A. L. Maniero, and U. Segre, Chem. Phys. 161, 257
29
S. Goshen, M. Friedman, R. Thieberger, and J. A. Weil, J. Chem. Phys. 79, (1992).
36
4363 (1983). A. R. Sørnes and N. P. Benetis, Chem. Phys. 226, 151 (1998).
30 37
E. Ley-Koo and S. Rubinstein, J. Chem. Phys. 73, 887 (1980). C. L. Gardner and E. J. Casey, Can. J. Chem. 46, 207 (1968).
31 38
K. Prassides, T. J. S. Dennis, C. Christides, E. Roduner, H. W. Kroto, M. Prager and A. Heidemann, Chem. Rev. 97, 2933 (1997).
R. Taylor, and D. R. M. Walton, J. Phys. Chem. 96, 10600 (1992).
32
W. Press, Single-Particle Rotations in Molecular Crystals, Springer Trends in
Modern Physics (Springer Verlag, Berlin 1981), Vol. 92. Translated by AIP Author Services

Low Temp. Phys. 45, 000000 (2019); doi: 10.1063/1.5093524 45, 000000-440
Published under license by AIP Publishing.