
2
dering, which transforms to a true phase for a quasi-two-
dimensional system. In contrast, the discrete symmetry
of the CDW allows for a true phase transition. In ad-
dition, technically speaking, in the present work, we are
performing calculations for finite systems, where long-
range fluctuations are eventually cut off, so neither the
AFM or CDW state are strictly speaking phases. Nev-
ertheless, in the following, we will refer to both states
as phases, since we are interested in the interplay of the
competing fluctuations corresponding to these orderings.
Our conclusions should thus be understood as applying
either to finite systems replacing the notion of phase by
”state dominated by the respective fluctuations” or to
a quasi-two-dimensional system in the thermodynamic
limit.
Limitations in the treatment of competing collective
fluctuations arise in the currently available approaches
employed for studying quantum lattice systems. Nu-
merically exact methods, such as exact diagonalization
(ED) [1] and lattice Monte Carlo [2] have studied the
interplay between Uand V[15,16,20,21] but are re-
stricted to small system sizes and thus cannot address
long-range collective fluctuations. The same problem is
also inherent in cluster extensions of the dynamical mean-
field theory (DMFT) [44–49], such as, e.g., DCA [38–
40]. Diagrammatic methods based on the parquet ap-
proximation [50–55] allow one to account for the inter-
play between charge and spin fluctuations [26] originat-
ing from the two-particle vertex functions in an unbi-
ased and powerful fashion. These vertices are incorpo-
rated with full momentum- and frequency-dependence,
and the approach is thus computationally very expen-
sive, which severally limits its applicability. Advanced
diagrammatic extensions of DMFT [56] are able to de-
scribe long-range fluctuations simultaneously in different
instability channels. In the presence of the non-local
interaction Vthis can be done within the dual boson
theory [34,36,57–59], the dynamical vertex approxi-
mation (DΓA) [60,61], the triply irreducible local ex-
pansion (TRILEX) method [62], or the dual TRILEX
(D-TRILEX) approach [27,28,63]. However, these fluc-
tuations are usually treated in a ladder-like approxi-
mation, where different instability channels affect each
other only indirectly via self-consistent renormalization
of single- and two-particle quantities.
Current approaches to quantum lattice systems that
are able to capture competing collective fluctuations are
too complicated for broad usage. In this work, we de-
velop a multi-channel generalisation of the fluctuating
field (FF) approach that allows us to incorporate multiple
collective fluctuation channels and their interplay in a nu-
merically cheap way without explicitly breaking the sym-
metry of the model. The FF method was originally intro-
duced for the study of spin fluctuations in the classical
Ising plaquettes [64] and was further developed for single-
and multi-mode treatment of collective spin fluctuations
in the Hubbard model [65–67]. We employ the pro-
posed multi-channel fluctuating field (MCFF) approach
to study the interplay between CDW and AFM fluctu-
ations in the extended Hubbard model on a half-filled
square lattice with a repulsive on-site Uand nearest-
neighbour Vinteractions. We show that the MCFF ap-
proach predicts results for the CDW and AFM phase
boundaries in qualitative agreement with more elaborate
numerical methods. Furthermore, it allows to model
competing collective fluctuations for large system sizes
near the thermodynamic limit. In addition, the method
is able to distinguish between stable and meta-stable col-
lective fluctuations. For this reason, the MCFF approach
allows us to capture the true ground state of the coex-
istence region of CDW and AFM fluctuation that was
obtained in Ref. [25] on the basis of DCA calculations.
II. MODEL
For simplicity, our considerations are limited to a
single-band extended Hubbard model. However, we note
that our approach can be straightforwardly generalised
to more complex single- and multi-band quantum lat-
tice systems. The Hamiltonian of the extended Hubbard
model has the following form:
ˆ
H=−tX
hi,ji,σ
ˆc†
iσ ˆcjσ +UX
i
ˆni↑ˆni↓+V
2X
hi,ji,σσ0
ˆniσ ˆnjσ0.
(1)
In this expression, ˆc(†)
iσ operators correspond to annihi-
lation (creation) of electrons, where the subscripts de-
note the position iand spin projection σ∈ {↑,↓}. Our
system is modelled by the hopping tbetween nearest-
neighbor sites hi, jion a two-dimensional square lat-
tice. The Coulomb interaction between electronic densi-
ties ˆniσ = ˆc†
iσ ˆciσ contains the on-site Uand the nearest-
neighbor Vcomponents.
The extended Hubbard model (1) displays two sym-
metries of fundamental importance for our considera-
tions: a continuous SU(2) symmetry associated with
spin degrees of freedom and a discrete particle-hole sym-
metry related to charge degrees of freedom. To facil-
itate our later treatments, we include a sketch of the
finite temperature U, V phase diagram of the extended
Hubbard model on the two-dimensional square lattice in
Fig. 1. Within the sketch, we denote the regime of strong
CDW fluctuations (red gradient), with asymptotics of
the CDW phase boundary highlighted, and the regime
of strong AFM fluctuations (blue gradient). The CDW
phase boundary occurs along V=U/8 + cst. at weak
coupling [33], which transforms to V=U/4 at interme-
diate coupling [20], followed by V∼U+ cst. at strong
coupling [29,34,36,57,68]. At weak coupling the AFM
phase boundary starts at a critical U, which further ex-
tends to the V=U/4 phase boundary at intermediate
coupling [25]. We restrict our consideration to the weak
to intermediate coupling regime, with the strong coupling
regime being outside the scope of the current work.