
2
dominates in absence of band carriers33,34.
The maximum total magnetic moment in
LaMn1−xGaxO3is achieved for x=0.5. For x>0.6
the lack of Jahn-Teller effect makes the system cubic20.
Unusually for ferromagnetic compounds, these samples
are also electrically insulators. The gallium doping
has dramatic effect at concentrations lower than 5%,
since one Ga atom increases the magnetic moment in
an applied magnetic field up to 16 µBper Ga atom23.
Investigating similar ferromagnetism in LaMn1−xScxO3,
it was shown that the ordering of the Mn3+ Jahn-
Teller distortion is not disrupted in the ab-plane for
any Sc concentration. This contrasts with the results
of the LaMn1−xGaxO3, where a regular MnO6is
found for x>0.5. Therefore, both LaMn0.5Sc0.5O3and
LaMn0.5Ga0.5O3show a similar ferromagnetic behavior
independently of the presence or not of the Jahn-Teller
distorted Mn3+, pointing to the Mn-sublattice dilution
as the main effect in driving ferromagnetism over the
local structural effects35.
Here, using first principles calculations, we study
the electronic, magnetic and orbital properties of the
LaMn1−xGaxO3focusing on the interplay between the
Jahn-Teller and the octahedral distortions in the antifer-
romagnetic phase. The paper is organized as follows: we
present the computational details for the ab-initio cal-
culations in Section II, while in Section III we show the
results from first principles studies, focusing on the oc-
tahedral distortions, the density of state (DOS) and the
orbital order. Finally, in Section IV we propose the pos-
sible origin of the experimentally detected ferromagnetic
phase.
II. COMPUTATIONAL DETAILS
We perform spin-polarized first-principles density func-
tional theory (DFT) calculations36 using the Quan-
tum Espresso program package37, the GGA exchange-
correlation functional of Perdew, Burke, and Ernzerhof38,
and the Vanderbilt ultrasoft pseudopotentials39 in which
the La(5s, 5p) and Mn(3s, 3p) semicore states are in-
cluded in the valence. We used a plane-wave energy cut-
off of 35 Ry and a Gaussian broadening of 0.01 Ry as in
the reference 40. These values for the plane-wave cutoff
and the Gaussian broadening are used in all calculations
presented in this chapter. A 10 ×10 ×10 k-point grid
is used in all DOS calculations, while a 8 ×8×8 grid is
used for the relaxation of the internal degrees of freedom.
We optimized the internal degrees of freedom by mini-
mizing the total energy to be less than 10−4Hartree and
the remaining forces are smaller than 10−3Hartree/Bohr,
while fixing the lattice parameter a,band cto the experi-
mental values19,24. After obtaining the DFT Bloch bands
within GGA, we use the occupation matrix to obtain the
orbital order parameter. For the DOS calculations, we
used a Gaussian broadening of 0.02 eV to have an accu-
rate measurement of the bandgap.
To go beyond GGA, it has been proposed to in-
clude the Coulomb repulsion Uinto the GGA theory
giving rise to the so-called GGA+Utheory. GGA+U
was first introduced by Anisimov and his coworkers41,42.
Here, we use the rotational invariant form introduced by
Lichtenstein43 in its spherically averaged and simplified
approach44 where there is just an adjustable parameter
Ueff =U−J. A self-consistent method for the deter-
mination of Ueff was proposed by Cococcioni et al.45
Starting from the observation of the non piecewise be-
haviour of the energy as a function of the occupation
number42, they implemented a method to take into ac-
count the electron screening in the Hubbard repulsion.
We used the refined approach suggested by Cococcioni
seeking internal consistency for the value of Uef f . Once
calculated the first value of Ueff within the GGA scheme,
we performed the Cococcioni technique for the functional
GGA+Ueff obtaining a correction to the Uef f value and
repeating the procedure until the correction for the fi-
nal value of Ueff vanishes. We constructed several su-
percells required by the Cococcioni method and we cal-
culated Ueff from the AFM configuration using exper-
imental volume and atomic positions46. The well con-
verged value for the AFM configuration of LaMnO3is
Ueff =6.3 eV. The Uef f obtained in the AFM phase is
in good agreement with Ueff values between 3.25 and
6.25 eV used for LaMnO36,47 and other Mn-based insu-
lating systems48–51. We use the value U=6.3 eV in all
the configurations and at all the doping concentrations
of LaMn1−xGaxO3.
The optimization of manganites within DFT is a very
delicate issue and no clear methodology (GGA, LSDA,
GGA+U, etc) has been assessed in the literature, each
having advantages and disadvantages, as for the descrip-
tion of tilting, JT distortion, lattice parameters etc. We
perform the relaxation of the atomic positions at fixed
experimental volume for some doping concentrations in
GGA in the AFM spin configuration using several su-
percells. After the relaxation, we calculate the density
of state, the distortion and the orbital order in GGA+U
using the Ucalculated by the Cococcioni method.
Here, we study the properties of the LaMn1−xGaxO3
at x=0.000, 0.125, 0.250, 0.500 considering a 2 ×2×2
supercell for x=0.125 and 0.250 while we consider a
√2×√2×2 supercell for the other cases. We substi-
tute the Ga atom to the Mn atom in the centre of the
octahedron. We use the experimental volume from 19
and 24 to construct the supercells. The net magnetic
moment of the supercell will be different from zero at
x=0.125, 0.250 due to the presence of an odd number of
Mn atoms. The impurities distribution could be treated
with more advances techniques to get accurate quantita-
tive results, however, the results obtained within these
approximations return satisfactory qualitative results.