2
tice dynamics for various systems with different bonding
characteristics[35].
In this article, we examine the role magnetoelastic ef-
fects play in explaining the experimental phonon disper-
sion of pristine YBa2Cu3O6by taking advantage of the
numerically stable r2SCAN functional. We find specific
branches of the phonon band structure to be sensitive
to the ground state magnetic order. Moreover, these
phonons correspond to breathing modes within the CuO2
plane, suggesting a sensitive dependence on magnetoe-
lastic coupling, which may facilitate a positive-feedback
loop between electronic, magnetic, and lattice degrees of
freedom.
II. METHODS
Ab initio calculations were performed using the pseu-
dopotential projector-augmented wave method[36, 37]
implemented in the Vienna ab initio simulation pack-
age (VASP) [38, 39] with an energy cutoff of 600 eV
for the plane-wave basis set. Exchange-correlation ef-
fects were treated using the r2SCAN [34, 35] meta-GGA
scheme. The calculations are performed with a Gamma-
centered mesh having a spacing threshold of 0.15 ˚
A−1
for the k-space sampling. We used the experimental
low-temperature P4/mmm crystal structure to initialize
our computations[40]. All atomic sites in the unit cell
along with the cell dimensions were relaxed using a con-
jugate gradient algorithm to minimize the energy with
an atomic force tolerance of 0.001 eV/˚
A and a total en-
ergy tolerance of 10−7eV. The harmonic force constants
were extracted from VASP using the finite displacement
method (with displacement 0.015˚
A) as implemented in
the Phonopy code [41]. In some calculations, to give bet-
ter agreement with the experimental volume an effective
U was added to r2SCAN[34].
III. RESULTS
Table I compares various properties calculated with the
non-magnetic (NM) and antiferromagnetic (AFM) states
to available experimental values for YBa2Cu3O6. In the
NM state, the lattice parameters and corresponding vol-
ume are the furthest away from the experimental values.
Specifically, the in-plane lattice parameters slightly un-
derestimate those from neutron scattering, whereas the
predicted c-height is significantly larger, consistent with
previous studies employing PBE and SCAN[18, 19].
When the majority and minority spins are allowed to
self-consistently relax, we stabilize the experimentally
observed G-type AFM order across the planar copper
atomic sites. The predicted value of the magnetic mo-
ment on copper sites is 0.45 µB, which is in good accord
with the corresponding experimental value of 0.55 µB[45].
Due to the localization of electrons on the copper sites
the ab-plane expands with a concomitant shrinking of the
c-axis, bringing the equilibrium crystal geometry in line
with the experimental values. Since the phonon disper-
sion is determined by the inter-atomic forces, which de-
pends sensitively on the ground state electronic structure
and equilibrium atomic positions, the excellent perfor-
mance of r2SCAN in predicting the equilibrium ground
state bodes well for an accurate prediction of the lattice
dynamics.
Figure 1 (a) compares the theoretically predicted
phonon dispersion of YBa2Cu3O6in the AFM phase
with the experimental bands obtained by inelastic neu-
tron scattering [42–44]. For convenience we plotted the
phonon spectra in the NM Brillouin zone of Fig. 1 (b).
Overall, r2SCAN yields phonon frequencies and their dis-
persion along all three directions in momentum space in
excellent accord with experiment.
To analyze the sensitivity of the phonon bands to
magnetoelastic coupling we compare the NM and AFM
phonon dispersion in Fig. 2 with the experimentally re-
ported bands overlaid. By inspection, it is clear that
a majority of the phonon branches are not significantly
affected by the change in electronic environment [NM
vs. AFM], which is consistent with the small difference
in lattice constants given in Table I. However, for some
branches, the AFM order results in a hardening of the af-
fected phonon branches, i.e. a shift to higher frequency,
which improves agreement with experiment.
While overall agreement between experiment and the
AFM phonon dispersions is quite good, for several
branches there remains an underestimation of the hard-
ness of the atomic bonds. Interestingly, many of these
branches seem to be of relevance for the electronic prop-
erties of the cuprates. In Fig. 2, we highlight seven
branches, A-D along Γ-X, and E-G along Γ-M, for special
discussion, illustrating the associated atomic displace-
ments in Fig. 2(b) and listing properties of the A, B,
D, and G modes in Table II. Not only do these branches
mostly show strong effects of magnetic order, but most
are also of experimental interest, having strong doping
dependence or changes at the superconducting transi-
tion. These branches are all related to the deformation
(stretching or buckling) of Cu-O bonds. Branches A and
B feature Cu-O bond out-of-plane buckling vibrations,
while branch F features buckling in the CuO2plane.
Branches C, D, E and G are all related to Cu-O bond
stretching, whereas branches C, D and G also have a no-
table admixture of apical Cu-O bond stretching. This
admixture has been noted in previous studies[46]. While
branch D is usually called a half-breathing mode con-
sidering only the motion of the in-plane oxygens, there
is a strong motion of the apical oxygen so we term this
branch as an ac-plane full breathing mode. Branch G,
normally called the full-breathing mode, is denoted as
the 3D full-breathing mode. Mode E is normally called
quadrupolar mode. Mode F is called the scissor mode
in terms of the in-plane Cu-O bond buckling behav-
ior. The Cu-O bond stretching modes are experimen-
tally interesting due to their softening upon doping [47],