Content uploaded by Peter Lemmens
Author content
All content in this area was uploaded by Peter Lemmens
Content may be subject to copyright.
arXiv:0904.0233v1 [cond-mat.str-el] 1 Apr 2009
Interplay of electronic correlations and lattice instabilities in
BaVS3
Kwang-Yong Choi,1,2Dirk Wulferding,2Helmuth Berger,3and Peter Lemmens2
Department of Physics, Chung-Ang University, 221 Huksuk-Dong,
Dongjak-Gu, Seoul 156-756, Republic of Korea
Institute for Condensed Matter Physics,
TU Braunschweig, D-38106 Braunschweig, Germany and
Institute de Physique de la Mati´ere Complexe,
EPFL, CH-1015 Lausanne, Switzerland
(Dated: April 1, 2009)
Abstract
The quasi-one-dimensional metallic system BaVS3with a metal-insulator transition at TM I =
70 K shows large changes of the optical phonon spectrum, a central peak, and an electronic Ra-
man scattering continuum that evolve in a three-step process. Motivated by the observation of a
strongly fluctuating precursor state at high temperatures and orbital ordering and a charge gap at
low temperatures we suggest a concerted action of the orbital, electronic, and lattice subsystems
dominated by electronic correlations.
PACS numbers: 71.27.+a, 71.30.+h, 78.30.-j
1
The discussion on the relation of lattice distortions and electronic correlations effects
has recently been refueled by the good agreement that certain correlated electron models
show in calculating structural relaxations of transition metal compounds [1]. This implies
in some sense a superiority of electronic with respect to lattice degrees of freedom. On
the other side, the considered system, KCuF3, is a high symmetry compound with a single
hole in an egstate. This realizes a comparatively simple and clear electronic and structural
situation. The lattice distortion, as a cooperative Jahn-Teller transition, happens at very
high temperatures and the related orbital orbital ordering leads to its ambient temperature,
low energy description [2] as a one dimensional spin s=1/2 chain system. However, correlated
electron systems with metal-insulator transitions (MIT) accompanied by Mott localization
and charge ordering [3] are wellknown for more complex electronic systems involving multiple
orbitals. They notoriously tend to complex structural distortions and a low symmetry
crystallographic ground states. It is therefore of capital importance to perform further
studies on the relation of structural and electronic degrees of freedom on such a more complex
correlated electron systems.
In this context, the quasi-one-dimensional (quasi-1D) metallic system BaVS3is a fas-
cinating example to study cooperative spin, charge, and orbital correlations with reduced
dimensionality, highlighted by the observation of a Luttinger-liquid behavior [3] and focussed
theoretical studies [4, 5]. In spite of extensive experimental and theoretical efforts on this
system, however, there is no consensus on the phase diagram and the hierarchies of the
involved energy scales [6].
In our Raman scattering study, we provide direct evidence for an orbital-ordered and
charge-gapped ground state of BaVS3. In addition, the existence of a strongly fluctuating
state in the metallic phase and a unstable state in the insulating phase suggests an intimate
coupling of the different subsystems.
BaVS3crystallizes in the hexagonal structure (P63/mmc with Z=2) at room tempera-
ture [7]. The VS3spin chains are formed by stacking face sharing VS6octahedra along the c
axis. Each chain is separated by Ba atoms in the ab planes, realizing a quasi-1D structure.
Two electrons in the 3d t2gatomic levels are split into a broad A1gband (dz2orbital) and
two quasidegenerate narrow e(t2g) bands in the trigonal crystal field. The former is respon-
sible for the 1D metallic behavior while the latter dictate localized 3D electrons [4, 8, 9, 10].
Recent angle resolved photoemission spectroscopy (ARPES) [11] and local density approx-
2
imation (LDA) combined with dynamical mean-field theory [4] have highlighted the sig-
nificance of large on-site Coulomb repulsion and Hund’s rule coupling leading to nearly
half-filling of both A1gand Eg1bands. With decreasing temperature BaVS3undergoes three
successive phase transitions: First, a second-order structural phase transition to the or-
thorhombic Cmc21space group at TS=240 K with distorted V4+ zigzag chains. Second,
a MIT at TM I =70 K accompanied by a symmetry lowering to a monoclinic phase (Im),
a sharp maximum of the magnetic susceptibility and specific heat anomalies [12]. Third,
magnetic ordering for T<TX= 30 K.
The main driving force of the MIT is proposed to be a Peierls-like transition stabilizing a
2kF(A1g) CDW with a critical wave vector qc=c∗/2, precipitated by huge precursor fluctu-
ations at about 170 K [13]. Remarkably, an internal distortion of the VS6octahedra and a
tetramerization of the V4+ chains occur without appreciable charge disproportionation [14].
This was discussed in terms of an orbital ordering in the insulating phase. The orbital-
ordering scenario was supported by several experimental investigations [14, 15, 16, 17] and
theoretical modelling [4] although direct evidence is still lacking.
Raman spectroscopy is a tool of choice to address these issues because optical phonons
and electronic scattering can serve as a local probe to explore the evolution of the orbital,
electronic and lattice subsystem. The used single crystals of BaVS3grown by the Tellurium
flux method [16] had typical dimensions of 3 ×0.2×0.2 mm3. Raman scattering experiments
were performed in quasi-backscattering using the λ= 488 nm excitation line with a power
of 5 mW focused on 0.1 mm diameter.
Figure 1 compares Raman spectra for polarization parallel (cc) and perpendicular (bb) to
the chain direction at 4 K and 280 K, respectively. In the high-temperature metallic phase
two peaks around 157 and 365 cm−1are superimposed by an electronic continuum. They
belong to four symmetry-allowed modes; ΓRaman = 1Ag(xx, yy, zz) + 3E2g(xx −yy, xy) [18].
The 157 cm−1E2gand 365 cm−1A1gmodes correspond to a vibration of the S atoms and a
breathing mode of the VS6octahedra, respectively. These modes are weak and broad due to
the screening by conduction electrons. In the low-temperature insulating phase we observe 8
rather broad A1modes with inter-chain polarization. This agrees well with the factor group
prediction ΓRaman = 8A1(xx, yy, zz) for the space group Cmc21. In in-chain polarization
weak satellite peaks are observed in addition to the sharp, intense principal modes. The
electronic background shows a drastic change as well. The markedly different polarization
3
FIG. 1: Raman spectra of BaVS3for light polarizations parallel (cc) and perpendicular (bb) to the
chain direction at 4 K and room temperature, respectively. The numbers denote the frequency of
principal phonon peaks at 4 K.
dependence corresponds to anisotropic electronic polarizabilities in the quasi-1D nature of
BaVS3.
Shown in Fig. 2 is the detailed temperature dependence of the Raman response Imχ,
which is corrected by the Bose thermal factor [1 + n(ω)] = [1 −exp(−¯hω/kBT)]−1from the
measured Raman scattering intensity. With decreasing temperature the electronic scattering
is suppressed and shows a depletion of spectral weights below ∼400 cm−1(= 49.6 meV) for
T < 40 K, indicative of the opening of a charge gap. The gap obtained from Raman
measurements falls in the same range as in optical conductivity (∆ch ∼42 meV) [19],
photoemission (∆ch = 60−70 meV) [11], and resistivity measurements (∆ch ∼52 meV) [16].
We note that in the ARPES and optical conductivity experiments a pseudogap feature is
visible up to ∼80 K. In contrast, the charge gap is discernible in our Raman spectra for
T < TX. This is because the pseudogap feature is covered by other low-energy excitations
above TX(see below).
The phonon modes exhibit a number of interesting and anomalous changes in frequency
and linewidth as a function of temperature. For a quantitative analysis we fit them to
4
FIG. 2: Temperature dependence of the Raman response Imχwith in-chain polarization. The
dashed lines are the scattering background. The data are systematically shifted and multiplied by
the given factor.
Lorentzian profiles. The resulting frequency, linewidth, and intensity of the 167, 206, 326,
and 385 cm−1modes are summarized in Fig. 3. The other modes have essentially the same
behavior (not shown here).
Upon cooling from TSto TM I the phonon energies undergo a giant energy shift by
10 −40 cm−1. Also their linewidths decrease enormously whereas their intensities show
no substantial change. Between TM I and TXthe phonon modes exhibit sizable changes in
both intensity and linewidth. In contrast, the phonon energies hardly vary with tempera-
ture. Below TXthe intensities increase abruptly and the linewidths drop rapidly. The most
salient feature is a splitting of the phonon modes. Two new peaks show up as a shoulder
of each principal mode with weaker intensities. In addition, all 8 A1modes have the same
three-peak feature. This suggests that the new modes are due to a zone folding of zone
boundary modes to the Γ point. The two zone-folded modes imply the presence of a new
periodicity with four atomic spacings known as a tetramerization of V4+ chains [14]. Since
the zone-folded modes are clearly resolved only with in-chain polarization they should be
5
FIG. 3: Temperature dependence of the peak frequencies (upper panel), the linewidth (middle
panel), and intensity (lower panel) of the 167, 206, 326, and 385 cm−1modes. The vertical dashed
lines denote the consecutive phase transitions.
ascribed to the 4kFeffects of the essentially 1D metallic bands. However, a conventional
CDW does not give rise to a backfolding of all phonon branches. Rather, our results are
compatible with a periodic orbital arrangement of the four V4+ atoms, which induces strong
variations of electronic polarizabilities. Therefore, we give spectroscopic evidence for a static
orbital ordering.
Next we will focus on the evolution of the background scattering. In the high temperature
metallic phase the electronic Raman response is well fitted by a collision-dominated (CD)
model, Imχ∝BωΓ/(ω2+ Γ2), where B is the scattering amplitude and Γ is the carrier
scattering rate. Approaching TSa broad elastic-scattering maximum becomes prominent,
which is described by a Gaussian profile. This is a central peak arising from the second-order
structural phase transition [20]. We find that in a wide temperature range of 50 < T < TS
the background response is well fitted by the sum of the CD scattering (electronic response)
and the Gaussian profile (cental peak). Representative fits are displayed in Fig. 4(a) with
6
FIG. 4: (a) Electronic Raman response at 280, 170, and 70 K obtained after subtracting phonon
peaks. The fitting corresponds to a collision-dominated model (dashed lines), a Gaussian profile
(dotted) and the sum of the two contributions (solid). (b) Temperature dependence of the scattering
amplitude (full squares) and full-width at half maximum (inverse open triangle) extracted from
the Gaussian fit. (c) Temperature dependence of the scattering amplitude obtained by a collision-
dominated model. (d) Temperature dependence of the intensity of the 385 cm−1zone-folded mode
(full square) and the linewidth of the 385 cm−1principal mode (open inverse triangle). Solid and
dashed lines are guides to the eye.
fitting parameters summarized in Figs. 4(b) and (c). The intensity of the central peak
becomes stronger upon cooling through TS, shows a maximum around 210 K and then
decreases rapidly with a small maximum at TM I , consistent with the second structural
phase transition. Both its linewidth and the amplitude of the electronic response resemble
the temperature dependence of the intensity. Surprisingly, no well-defined soft mode shows
up below TS. Instead, the central peak persists well below TSwith an exceptionally broad
full-width at half maximum (FWHM) of several hundreds cm−1. This feature indicates the
presence of electronic and structural fluctuations in nearly the whole temperature range.
In the following we will reflect on the underlying mechanism of the MIT. The high-
temperature metallic state (T > TS) is distinguished from simple metals due to the CD
scattering with a carrier scattering rate Γ = 230 cm−1. The metallic A1gelectrons alone
cannot yield such a low-energy electronic Raman response due to screening by itinerant
electrons [21]. Therefore, the presence of the CD response means that there exist strong
7
electronic scattering from spin(orbital) fluctuations of the localized Egelectrons. This signals
that electronic correlations are a driving factor of the manifold MIT.
In the intermediate-temperature metallic phase (TM I < T < TS) there appears a strong
renormalization of the phonon energies, a three-times decrease of the phonon linewidths, and
a suppression of the central peak and CD response with decreasing temperature. In our case,
the phonon linewidth is determined by an electron-phonon mechanism and thus is a measure
of the free carrier concentration. Our observation indicates a continuous reduction of the
electronic density of states at the Fermi level EF, which diminishes the central peak and
CD scattering amplitudes [compare the middle panel of Fig. 3 with Figs. 4(b) and (c)]. Our
results unveil that an electronic instability starts around TSand develops gradually. This
is fully consistent with the persistence of the fluctuating 2kF(A1g) CDW superstructure
up to 170 K [13]. The evolution of the central peak provides complimentary information
about the lattice instability because it is associated with the decay of a soft mode into
acoustic phonons or phonon density fluctuations. The enduring large linewidth down to
50 K indicates persisting structural fluctuations. Together with the similar temperature
dependence of the CD response due to localized Egelectrons we conclude that the Fermi
surface, structural (orbital), and lattice instabilities develop consonantly and their strong
fluctuations characterize the precursor transition regime.
In the low-temperature insulating phase (TMI < T < TX) there is no shift of the phonon
energies. In contrast, the phonon linewidths decrease moderately and the central peak and
CD scattering amplitudes drops rapidly just after reaching a maximum at TM I . The zone-
folded modes related to a 2kF(A1g) CDW instability is not seen. This is consistent with
the lack of a magnetic long-range order between TM I and TS[16]. Through the MIT the
orbital, electronic, and structural states become quasi-static but not totally frozen-in. In
this regime, several metastable states compete with each other.
For T < TX, there appear the zone-folded modes associated with the 4kFinstability, which
is a measure of the orbital order parameter. Their intensity increases strongly upon cooling
from 40 K [see Fig. 4(d)]. At the same time, the cental peak and CD electronic scattering
disappear. Instead, the deletion of the low-frequency electronic scattering evidences the
opening of a charge gap. This is further supported by an exponential-like drop of the
linewidth, which means a depletion of the density of states at EF. The fact that the onset
temperature of the superstructure peak and the drop of the linewidth coincide corroborates
8
that the orbitals are important in stabilizing the charge-gapped insulating ground state.
To conclude, we demonstrate that in BaVS3electronic fluctuation exist already above
the phase transitions. We find that the exceptionally rich phases are initiated by electronic
correlations. These effects in turn cause the entangled instability of the orbital, electronic,
and lattice subsystem. Thus, the ”order parameter” of the MIT in this compound is a
complex quantity involving all three subsystems.
This work was supported by the DFG, the ESF program Highly Frustrated Magnetism,
the Swiss NSF and the NCCR MaNEP. KYC acknowledges financial support from the
Alexander-von-Humboldt Foundation.
[1] I. Leonov et. al., Phys. Rev. Lett. 101, 096405 (2008).
[2] See, for example, P. Fulde, Electron Correlations in Molecules and Solids, (Springer-Verlag,
Berlin, 1995).
[3] M. Imada et. al., Rev. Mod. Phys. 70, 1039 (1998).
[4] F. Lechermann et. al., Phys. Rev. B 76, 085101 (2007) and references therein.
[5] K. Penc et. al., Phys. Rev. B 68, 012408 (2003).
[6] P. Fazekas et. al., Journ. Magn. Magn. Mat. (2009) and cond-mat/0702510.
[7] R. Gardner et. al., Acta Crystallogr., Sect. B: Struct. Crystallogr. Cryst. Chem. 25, 781
(1969).
[8] M.-H. Whangbo et. al., J. Solid State Chem. 165, 345 (2003).
[9] F. Lechermann et. al., Phys. Rev. Lett. 94, 166402 (2005).
[10] F. Lechermann et. al., Phys. Rev. B 74, 125120 (2006).
[11] S. Mitrovic et. al., Phys. Rev. B 75, 153103 (2007).
[12] H. Imai et. al., J. Phys. Soc. Jpn. 15, 3460 (1996).
[13] S. Fagot et. al., Phys. Rev. Lett. 90, 196401 (2003).
[14] S. Fagot et. al., Phys. Rev. B 73, 033102 (2006).
[15] H. Nakamura et. al., Phys. Rev. Lett. 79, 3779 (1997).
[16] G. Mih´aly et. al., Phys. Rev. B 61, R7831 (2000).
[17] T. Ivek et. al., Phys. Rev. B 78, 035110 (2008).
[18] Z. V. Popovic et. al., Phys. Rev. B 65, 132301 (2002).
9
[19] I. K´ezsm´arki et. al., Phys. Rev. Lett. 96, 186402 (2006).
[20] K.-Y. Choi et. al., Phys. Rev. B 68, 104418 (2003) and references therein.
[21] A. Zawadowski and M. Cardona, Phys. Rev. B 42, 10732 (1990).
10