PHYSICAL REVIEW B 105, 024510 (2022)

Enhancing d-wave superconductivity with nearest-neighbor attraction

in the extended Hubbard model

Mi Jiang
Institute of Theoretical and Applied Physics, Jiangsu Key Laboratory of Thin Films, School of Physical Science and Technology,
Soochow University, Suzhou 215006, China

(Received 1 September 2021; revised 8 December 2021; accepted 5 January 2022; published 13 January 2022)

Motivated by the recent discovery of the anomalously nearest-neighbor attraction arising from the electron-
phonon coupling, we quantitatively investigate the enhancing effects of this additional attractive channel
on the d-wave superconductivity based on dynamic cluster quantum Monte Carlo calculations of a doped
two-dimensional extended Hubbard model with nearest-neighbor attraction −V . Focusing on the range of
0 < −V/t  2, our simulations indicate that the dynamics of d-wave projected pairing interaction is attractive at
all frequencies and increases with |V |. Moreover, turning on −V attraction enhances the (π , π ) spin fluctuations
but only enhances (suppresses) the charge fluctuations for small (large) momentum transfer. Thus, at V/t = −1
relevant to the “holon folding branch”, the charge fluctuations are insufficient to compete with the d-wave pairing
interaction strengthened by enhanced spin fluctuations. Our work suggests the underlying rich interplay between
the spin and charge fluctuations in giving rise to the superconducting properties.

DOI: 10.1103/PhysRevB.105.024510

I. INTRODUCTION Given the structural similarity among the cuprates,

the physics with nearest-neighbor attraction of the one-
The pairing mechanism mediated by virtual exchange
dimensional material should be naturally extended to two-
of a bosonic mode plays the key role in overcoming the
dimensional CuO2 planes. Because how to enhance the
Coulomb repulsion between electrons in order to give rise
superconducting Tc is an important open question, the effects
to a net attractive interaction for Cooper pairing. In con-
of the additional attractive channel and its interplay with the
ventional Bardeen-Cooper-Schrieffer superconductors, this
pure electron-electron interaction deserves more systematic
bosonic mechanism is realized by the retardation nature of
the electron-phonon interaction [1]. Despite that there is no exploration. Here we adopt an extended Hubbard model with
general consensus, there has been strong evidence that in both strong local repulsion and nearest-neighbor attraction
strongly correlated superconductors such as the cuprates and as the minimal model. In particular, we focus on the explicit
heavy fermion materials, the antiferromagnetic spin fluctua- enhancement of the d-wave SC by the inclusion of additional
tions, namely the magnons, play the role of the bosonic mode. strong nearest-neighbor attraction. The Hamiltonian reads as
 
In this scenario, the minimization of the repulsive interaction H =−t †
(ciσ c jσ + h.c.) + U ni↑ ni↓
due to the local Coulomb repulsion can be accomplished via i j,σ i
the sign changing of the pairing wave function, for instance, 
the dx2 −y2 -wave pair state in the cuprates [2]. +V niσ n jσ  (1)
Regarding the pairing mechanism in cuprates, there has i j,σ σ 
been long debate on the role of the electron-phonon in-
with the usual nearest-neighbor hopping t = 1 as the unit
teraction and particularly its relation to superconductivity
(SC). Although it is widely believed that the pure electron- energy scale, the on-site Coulomb repulsion U , and an addi-
electron interaction dominantly drive the Cooper pairing and tional nearest-neighbor Coulomb attraction V < 0. Note that
the electron-phonon coupling (EPC) only plays a minor role, this attractive V has an important difference from the con-
there has been spectroscopic evidence that the effects of strong ventional extended Hubbard model with repulsive V , which
electronic interaction and the EPC reinforce each other to has been widely studied for the physics induced by the nonlo-
drive a stronger SC in the strange-metal regime of Bi-2212 cal Coulomb repulsion [6–14]. Regarding its superconducting
[3], which indicates the possible enhancement of SC through properties, the consensus is that the d-wave pairing and the as-
multiple channels; for instance, the contribution from the sociated transition temperature are only weakly suppressed as
phonon coupling, in addition to the pure electronic interaction. long as the repulsive V does not exceed U/2. This robustness
In fact, the EPC does not only manifest its importance in the is owing to the retarded nature of d-wave pairing to minimize
enhancement of Tc . Most recently, comparative spectroscopic the impact of nonlocal repulsion [6,8,11]. In the case of neg-
and theoretical investigation of a one-dimensional cuprate ative V , it is naively expected that attractive nearest-neighbor
Ba2−x Sr x CuO3+d over a wide range of hole doping revealed interactions always enhance the SC because the neighboring
the existence of an anomalously strong nearest-neighbor attraction naturally contributes the d-wave pairs as indicated
attraction [4], which probably originates from the EPC, in ac- by early Hartree-Fock calculations [15]. Conversely, the re-
counting for the so-called “holon folding branch” feature [5]. cent numerical exact diagonalization study [7] uncovered that

2469-9950/2022/105(2)/024510(6) 024510-1 ©2022 American Physical Society

MI JIANG PHYSICAL REVIEW B 105, 024510 (2022)

the nearest-neighbor attractions also have thresholds above with conventional notation K = (K, iωn ), K  = (K , iωn ),
which the SC will be finally suppressed, which corrects the in- q = (q, iν) and the time-ordering operator T can be cal-
tuition that attractive and repulsive interactions have definitely culated numerically via a DCA cluster solver (CT-AUX in
opposite effects on SC. We emphasize that the enhanced SC our case). Then the cluster two-particle irreducible vertex
explored in this work is around the moderate 0  |V |  2t 
cσ σ  (q, K, K ) can be extracted through the Bethe-Salpeter
range, which is much smaller than the threshold needed to equation (BSE)
suppress SC, to be consistent with the amplitude of the anoma-
lous nearest-neighbor attraction |V | ∼ t extracted from both χcσ σ  (q, K, K  ) =χcσ
0  
σ  (q, K, K ) + χcσ σ  (q, K, K )
0


experimental and theoretical studies [4,5]. Also, we neglect × cσ  σ  (q, K , K  )χcσ  σ  (q, K  , K  ),
the important but still open question of whether the pure
(3)
Hubbard model at V = 0 hosts a superconducting ground state
or not [16,17]. 
where χcσ
0
σ  (q, K, K ) is the noninteracting two-particle
Green’s function constructed from the product of a pair of
fully dressed single-particle Green’s functions. The usual con-
II. DYNAMICAL CLUSTER APPROXIMATION
vention that the summation is to be made for repeated indices
Here we adopt the dynamical cluster approximation (DCA) is adopted.
[18–20] with a continuous time auxilary field (CT-AUX) Note that the above formalism Eqs. (2) and (3) has their
quantum Monte Carlo (QMC) cluster solver [21] to nu- counterparts for the corresponding lattice quantities, whose
merically solve the model Eq. (1). As one of various numerical calculations are, however, impractical due to their
embedded-cluster methods, similar to the cluster dynamical continuous nature. Therefore, one of the key DCA assump-
mean field theory (cDMFT), DCA maps the bulk lattice prob- tions is that the cluster two-particle irreducible vertex c is
lem onto a finite cluster of size Nc , whose physics involving used as the approximation of the desired lattice two-particle
complex interactions is solved exactly by various methods e.g. irreducible vertex .
QMC and exact diagonalization, while the remaining degrees The two-particle irreducible vertex and associated BSE
of freedom are treated at the mean-field level. Precisely, the Eq. (3) can be classified according to the superconducting,
first Brillouin zone is divided into Nc patches denoted by its charge, and magnetic channels. In this work, we are mostly
center wave vector K surrounded by N/Nc lattice wave vectors interested in the particle-particle superconducting channel for
k’s. In this way, the original lattice problem of N sites is the zero center-of-mass and energy. To this aim, the super-
simplified to an effective Nc -site cluster problem by coarse conductivity can be quantitatively displayed by the leading
graining the lattice single-particle Green’s function, which is eigenvalues of the BSE in the particle-particle channel in the
designed to converge to a cluster Green’s function obtained eigenequation form [25,26]
by the cluster solver mentioned earlier [18,20]. Although the T  pp
inter-cluster interactions can be treated more accurately with − (K, K  )χ̄0pp (K  )φα (K  ) = λα (T )φα (K ), (4)
Nc K 
an additional bosonic dynamic mean-field [22] as adopted in
the extended DMFT [23], in this work we neglect its dynamic where pp (K, K  ) denotes the irreducible particle-particle ver-
contribution for simplicity [11]. tex of the effective cluster problem with the cluster momenta
To achieve the goal of simulating a wide range of doping K and Matsubara frequencies ωn = (2n + 1)π T . Note that
levels, most of our calculations are for smallest Nc = 2 × 2 the spin indices are neglected for simplicity. Also, for the
DCA cluster to manage the sign problem of the underlying superconducting channel, q = (q, iν) = 0 is assumed since
CT-AUX QMC solver [21,24] down to the SC transition tem- our focus in this work is the even-frequency even-parity (spin
peratures T ∼ Tc . Despite the small cluster size, the pairing singlet) d-wave pairing tendency [25,26]. The coarse-grained
interaction and dynamics should be fully descriptive at this bare particle-particle susceptibility
level. In fact, the simulations with larger cluster Nc = 4 × 4
are also performed to (1) confirm the enhancing effects of the Nc 
χ̄0pp (K ) = G(K + k  )G(−K − k  ) (5)
attractive V while at higher temperature scale due to the QMC N k
sign problem and (2) to investigate the competing role of spin
and charge fluctuations in a finer momentum resolution. is obtained via the dressed single-particle Green’s function
To investigate the superconducting, charge, and magnetic G(k) ≡ G(k, iωn ) = [iωn + μ − εk − (K, iωn )]−1 , where k
instability of a particular model Hamiltonian, one has to de- belongs to the DCA patch surrounding the cluster momen-
termine the structure of the interaction responsible for these tum K, μ is the chemical potential, εk = −2t (cos kx + cos ky )
channels. Essentially, the cluster two-particle Green’s func- is the dispersion relation, and (K, iωn ) is the cluster self-
tion energy. In practice, we usually choose 16 discrete points
for both the positive and negative fermionic Matsubara fre-
 β  β  β  β quency ωn = (2n + 1)π T mesh for measuring the four-point

χcσ σ  (q, K, K ) = dτ1 dτ2 dτ3 dτ4 quantities like two-particle Green’s functions and irreducible
0 0 0 0
vertices. Therefore, the BSE Eq. (4) reduces to an eigenvalue
× ei[(ωn +ν)τ1 −ωn τ2 +ωn τ3 −(ωn +ν)τ4 ] problem of a matrix of size (32Nc ) × (32Nc ).
The eigenvalue λα (T ) gives the pairing tendency of
×T cK+q,σ
†
(τ1 )cKσ (τ2 )cK†  σ  (τ3 )cK +q,σ  (τ4 )
the superconducting channel; while the symmetry of the
(2) corresponding superconducting state is manifested by the

024510-2
ENHANCING D-WAVE SUPERCONDUCTIVITY WITH … PHYSICAL REVIEW B 105, 024510 (2022)

momentum and frequency dependence of the eigenvector

φα (K, iωn ). Note that the magnitude of λα (T ) denotes the
strength of the normal state pairing correlations. Accordingly,
the spatial, frequency, and more generally orbital dependence
of the eigenvector φα (K, iωn ) can be viewed as the normal
state analog of the superconducting gap to reflect the structure
of the pairing interaction [25,26]. The superconducting Tc is
extracted via the temperature where the leading eigenvalue
of Eq. (4) λ(Tc ) = 1. As expected for the extended Hubbard
model, the leading pairing symmetry occurs for the d-wave
channel with momentum structure cos Kx − cos Ky so that we
are only concerned in the leading eigenvalues λd and associ-
ated φd (K, iωn ).
As discussed by Scalapino [26], the two-particle irre-
ducible particle-particle vertex pp as the pairing interaction
is connected to the four-point vertex function, which also
contains information about the irreducible particle-hole ver-
tex ph in magnetic and charge channels. Thus, the pairing
interaction pp has an intrinsic relation to these particle-hole
channels ph . In fact, the dominant contribution of the d-wave
pairing interaction has been shown to arise from the spin-
one (S = 1) particle-hole exchange [25,26]. Therefore, in this
work we also extract the irreducible particle-hole vertex ph
in the magnetic and charge channels of the effective cluster
problem, respectively, from Eq. (3) and thereby we have the
BSE in the eigenequation form similar to Eq. (4) but with
coarse-grained bare particle-hole susceptibility
Nc 
χ̄0ph (q, K, K  ) = δKK  G(K + k  )G(K + k  + q). (6)
N k

The corresponding eigenvalues for the particle-hole channels

reflect the magnetic and charge instabilities, e.g., spin and
charge density waves.
In this work we are only interested in the case of zero
frequency transfer (iν = 0) similar to the particle-particle su-
perconducting channel Eq. (4). Note, however, that we keep
the momentum transfer to calculate q-dependent lattice sus-
ceptibilities, which can be obtained by the coarse-grained
two-particle Green’s function χ̄ ph (q, K, K  ) (instead of cluster
quantities that result in cluster susceptibilities), which is in
turn calculated via the coarse-grained BSE transformed from
Eq. (3) as
 −1
[χ̄ ph (q, K, K  )]−1 = χ̄0ph (q, K, K  ) − ph (q, K, K  ). (7) FIG. 1. (a) Temperature dependence of the leading (dx2 −y2 -wave)
Then our interested magnetic (s) and charge (c) lattice suscep- eigenvalue λd (T ) of BSE Eq. (4) in the particle-particle channel at
tibilities χs,c (q, T ) can be deduced as U/t = 7 and n = 0.9; (b) The filling ρ dependence of the d-wave
superconducting Tc extracted from λd (Tc ) = 1 reveals the enhancing
T 2  ph effect of nearest-neighbor attraction; (c) Comparison of the leading
χs,c (q, T ) = χ̄ (q, K, K  ). (8)
Nc2 K,K  eigenvalues for d-wave superconducting, q = (π , π ) antiferromag-
netic, and q = (0, 0) charge channels.
We refer the readers to Ref. [27] for more details of the DCA
formalism of the calculations of two-particle quantities.
clearly show the enhancing effects of V , Fig. 1(b) displays
the dependence of Tc extracted via λd (Tc ) = 1 on the filling.
III. RESULTS
One can see that approximately 10–15% enhancement of Tc
We first illustrate the temperature dependence of the lead- with finite attraction is a general feature for all fillings con-
ing d-wave eigenvalue λd (T ) for different V at fixed filling sidered here. Compared with the impact of repulsive V on
ρ = 0.9 in Fig. 1(a). Apparently, the nearest-neighbor at- the d-wave pairing [11], the variation of Tc with ±V is not
tractive V leads to the increase of λd (T ) as the evidence exactly symmetric over V = 0 but has roughly the same scale.
shows that the d-wave pairing tendency can be enhanced. To Therefore, in this sense, the d-wave pairing has similar robust-

024510-3
MI JIANG PHYSICAL REVIEW B 105, 024510 (2022)

FIG. 2. The d-wave projected irreducible particle-particle vertex FIG. 3. The leading d-wave eigenfunction φd (K = (π , 0), iωn )
for different attraction V at U/t = 7 and n = 0.9 at T /t =
d (iωm ) of BSE Eq. (4) for different V at U/t = 7, T /t = 0.1 and n = 0.9.
0.1. d is attractive at all frequencies. The retardation of φd becomes stronger with increasing |V |.

ness against the additional channel of attractive interaction

from nearest-neighbor −V . In other words, the anomalously spin configuration. At high frequencies, pp approaches the
nearest-neighbor attraction |V | ∼ t extracted from both ex- bare interaction V (q = K − K ), which is the Fourier trans-
perimental and theoretical studies to account for the “holon form of the nearest-neighbor interaction V . As confirmed in
folding branch” [4,5] does not have significant effects on the Fig. 2, for our 2 × 2 cluster, we have d (iωm ) ∼ 4V at large
d-wave superconducting Tc . iωm . Different from the repulsive V cases, where d (iωm ) is
Besides the d-wave superconductivity, the (extended) Hub- only attractive at low frequencies but repulsive at high enough
bard model can support other instabilities like spin and charge frequencies so that the effective attraction arises from the
density waves [6–14]. In particular, the additional nearest- low frequency components of d (iωm ) [11], the persistent
neighbor attraction is natural to host the charge ordering attractive nature of d (iωm ) reflects the cooperation between
instability. Therefore, to explore these instability apart from the original effective attraction from pure local Coulomb re-
superconductivity, Fig. 1(c) compares the temperature evolu- pulsion and the additional nearest-neighbor attraction.
tion of the leading eigenvalues for d-wave superconducting, The attractive feature and retardation nature of the d-wave
q = (π , π ) antiferromagnetic, and q = (0, 0) charge chan- pairing interaction can be reflected via the leading d-wave
nels. The choice of these two particular q’s is motivated by eigenvector φd (K, iωn ) of Eq. (2), whose frequency depen-
Fig. 4, which indicates that the leading lattice magnetic and dence is shown in Fig. 3 at K = (π , 0) and T = 0.1 for
charge susceptibilities located at these two specific momen- varying V . For all cases, φd falls to zero with a retardation
tum transfers. characteristic frequency scale, which mirrors the pairing in-
On the one hand, apparently, the dominant instability at teraction in Fig. 2.
low enough temperature is d-wave superconducting while at Although the linear change of d (iωm ) in Fig. 2 looks iωm
higher temperatures the antiferromagnetic ordering instability independent, the decisive factor in the BSE Eq. (4) is the ratio
exceeds the pairing one [25]. On the other hand, the expected between d at different V , which indeed strongly depends on
charge ordering instability is always the subleading one since the bosonic Matsubara frequency iωm . Also, the BSE also
our nearest-neighbor attraction V/t = −1 is still weak to sup- involves the coarse-grained bare two-particle susceptibility
press the leading d-wave pairing instability to induce the χ̄0pp , whose ratio between the values at different V is iωn
desired charge ordering. dependent as well. Therefore, it is not surprising that the
To have a better understanding of the pairing interaction variation of eigenvectors φd with V strongly depend on iωn
and its variation with the additional nearest-neighbor attrac- instead of simple linear change. In this mathematical sense,
tion, we resort to the d-wave projected dynamical pairing the relation between d and φd can be complex. However, it
interaction is physically plausible that φd becomes more retarded because
 of the additional nearest-neighbor attraction, which is similar
K,K gd (K)
pp
(K, iωn , K , iωn )gd (K ) to the phonon mediation induced retardation in conventional
d (iωm ) =  , (9)
2 superconductors. This is confirmed by the gradually increas-
K gd (K)
ing frequency scale of φd ’s decaying. This might also hint that
and its dependence on the bosonic Matsubara frequency ωm = at sufficiently strong attractive V , the leading eigenvector may
ωn − ωn , where gd (K) = cos Kx − cos Ky gives the d-wave lose the d-wave character, namely that the d-wave SC would
projection factor, and the fixed ωn = π T is adopted. As be finally destroyed and replaced by the competing charge
shown in Fig. 2, the pairing interaction d (iωm ) is attractive orders [7].
(negative) in all cases and the finite additional −V naturally As mentioned earlier, the pairing interaction pp has intrin-
strengthens the attractive interaction favoring neighboring sic relation to the irreducible particle-hole vertex ph . Given

024510-4
ENHANCING D-WAVE SUPERCONDUCTIVITY WITH … PHYSICAL REVIEW B 105, 024510 (2022)

and in turn push up Tc . Hence, the physical picture in terms

of the mediating role of spin fluctuations in SC is the same as
the system without V . In addition, owing to the additional at-
traction, the nearest-neighbor spin configurations are favored
to be compatible with the d-wave SC.
Compared with the behavior of χs , the charge susceptibil-
ity χc shows more nontrivial features. In particular, turning
on −V attraction enhances the charge fluctuation at small
momentum transfer, e.g., q = (0, 0), (π /2, 0) instead of q =
(π , π ) expected for repulsive V interaction [11]. The inset
of Fig. 4(b) at Nc = 4, T /t = 0.1 has a similar variation
with q. The favored nearest-neighboring charges are prone
to enhance the small q charge fluctuations, which coexists
with the d-wave pairing favored by q = (π , π ) spin fluctu-
ations. Therefore, both attractive and repulsive interactions
favor charge fluctuations but at different wave vectors, which
is reminiscent of the previous exact diagonalization investi-
gation based on the spin and charge structure factors of the
extended Hubbard model [7]. Apparently, as |V | exceeds some
threshold, the charge fluctuations would exceed the magnetic
fluctuations and finally destroy the d-wave SC. Summarizing
Fig. 4, at our interested moderate V/t = −1, the enhanced
SC originates from the enhanced spin fluctuation at large
momentum transfer q = (π , π ), where the associated charge
fluctuations are suppressed, namely the charge ordering ten-
dency at V/t = −1 is insufficient to suppress SC, whose
impact can only manifest itself at much large V attraction to
host the charge order at small q.
FIG. 4. The momentum transfer q dependence of the DCA lat-
tice (a) spin and (b) charge susceptibilities, χs and χc respectively, IV. SUMMARY
for DCA cluster Nc = 4 × 4, T /t = 0.3, U/t = 7, n = 0.9. Turn-
In conclusion, we adopted dynamic cluster quantum Monte
ing on −V attraction slightly (because of relatively high temperature)
enhances the spin fluctuations at all q’s but only enhances (sup-
Carlo calculations of the extended Hubbard model with
presses) the charge fluctuations for small (large) q. The inset shows nearest-neighbor attraction to study the impact of the addi-
the results for smaller cluster Nc = 4 but lower T /t = 0.1 to illustrate tional attractive channel on the d-wave SC. In particular, we
the enhancement of χs (π , π ). focus on the attractive interaction with amplitude |V | ∼ t,
which is motivated by the recent discovery of an anomalously
strong nearest-neighbor attraction probably arising from the
that the magnetic channel plays the central role in mediating electron-phonon couplings [4,5].
the d-wave pairing and also the additional nearest-neighbor It is found that the additional −V enhances the d-wave SC
interactions, either repulsive [6,8,11] or attractive [7], favor and the variation of Tc with ±V has roughly the same scale,
charge ordering, we calculate the zero frequency DCA lattice which confirms the expectation that the repulsive (attractive)
spin (s) and charge (c) susceptibilities via Eq. (8), whose V suppresses (enhances) the SC before it is ultimately de-
dependence on the momentum transfer q are shown in the stroyed by sufficiently large |V | [7]. Distinct from the case
main part of Fig. 4 for DCA cluster Nc = 4 × 4 and T /t = of repulsive V , the d-wave projected pairing interaction d
0.3, n = 0.9. At V = 0, the magnetic susceptibility χs peaks is attractive at all frequencies and its amplitude increases
at q = (π , π ) as expected for the repulsive Hubbard model with |V | and thereby favors the d-wave SC. Reflecting the
on square lattice, which is consistent with the scenario that behavior of d , the d-wave eigenfunction φd of the BSE in
the antiferromagnetic fluctuations mediate the d-wave pairing the particle-particle channel falls to zero with a characteris-
[25,26]. In turning on V/t = −1, χs exhibits a tiny increase tic frequency scale, which increases with |V | indicating the
at all q’s, which is due to the relatively high temperature stronger retardation induced by the additional −V . Further-
T /t = 0.3 to compromise with the severe sign problem at more, the examination of the momentum transfer q-resolved
lower temperature for large DCA cluster Nc = 4 × 4. The spin and charge susceptibilities indicates that, on the one
complementary inset of Fig. 4(a) explicitly shows the increase hand, (π , π ) spin fluctuations become stronger at V/t = −1
of χs at q = (π , π ), which is consistent with the increase compared with the case at V = 0; on the other hand, −V
of the eigenvalues in the magnetic channel [green line in attraction only enhances the charge fluctuations for small
Fig. 1(c)]. Apparently, the common peak structure of χs at momentum transfer instead of large wave vector, e.g., (π , π ).
finite V , namely the spin fluctuation is strongest at large mo- Thus, at the stage of our interested V/t = −1 relevant to the
mentum transfer, can enhance the d-wave pairing interaction “holon folding branch” [4,5], the charge fluctuations of much

024510-5
MI JIANG PHYSICAL REVIEW B 105, 024510 (2022)

smaller amplitude than its magnetic counterpart are insuffi- ACKNOWLEDGMENTS

cient to compete with the d-wave pairing interaction.
We acknowledge useful discussion with Z. Shi on the
Our presented work provides complemental knowledge
cuprates and the assistance of P. Mai for the code testing of
on the extensively studied extended Hubbard model, espe-
cially on quantitative examination of the role of the additional related projects. This work was supported by the startup fund
nearest-neighbor attraction uncovered recently. These results from Soochow University and Priority Academic Program
suggest the underlying rich interplay between the spin and Development (PAPD) of Jiangsu Higher Education Institu-
charge fluctuations in giving rise to the superconducting tions.
properties.

[1] P. Morel and P. W. Anderson, Phys. Rev. 125, 1263 (1962). [15] R. Micnas, J. Ranninger, S. Robaszkiewicz, and S. Tabor, Phys.
[2] D. J. Scalapino, Rev. Mod. Phys. 84, 1383 (2012). Rev. B 37, 9410 (1988).
[3] Y. He, M. Hashimoto, D. Song, S.-D. Chen, J. He, I. M. Vishik, [16] M. Qin, C.-M. Chung, H. Shi, E. Vitali, C. Hubig, U.
B. Moritz, D.-H. Lee, N. Nagaosa, J. Zaanen, T. P. Devereaux, Schollwöck, S. R. White, and S. Zhang, Phys. Rev. X 10,
Y. Yoshida, H. Eisaki, D. H. Lu, and Z.-X. Shen, Science 362, 031016 (2020).
62 (2018). [17] M. Qin, T. Schafer, S. Andergassen, P. Corboz, and E. Gull,
[4] Z. Chen, Y. Wang, S. N. Rebec, T. Jia, M. Hashimoto, D. Lu, B. arXiv:2104.00064.
Moritz, R. G. Moore, T. P. Devereaux, and Z.-X. Shen, Science [18] T. Maier, M. Jarrell, T. Prushke, and M. Hettler, Rev. Mod.
373, 1235 (2021). Phys. 77, 1027 (2005).
[5] Y. Wang, Z. Chen, T. Shi, B. Moritz, Z.-X. Shen, and T. P. [19] U. R. Hahner, G. Alvarez, T. A. Maier, R. Solca, P. Staar, M. S.
Devereaux, Phys. Rev. Lett. 127, 197003 (2021). Summers, and T. C. Schulthess, Comput. Phys. Commun. 246,
[6] D. Sénéchal, A. G. R. Day, V. Bouliane, and A.-M. S. Tremblay, 106709 (2020); The DCA + + code used for this project can be
Phys. Rev. B 87, 075123 (2013). obtained at https://github.com/CompFUSE/DCA.
[7] N. Plonka, C. J. Jia, Y. Wang, B. Moritz, and T. P. Devereaux, [20] M. H. Hettler, A. N. Tahvildar-Zadeh, M. Jarrell, T. Pruschke,
Phys. Rev. B 92, 024503 (2015). and H. R. Krishnamurthy, Phys. Rev. B 58, R7475(R) (1998).
[8] A. Reymbaut, M. Charlebois, M. F. Asiani, L. Fratino, P. [21] E. Gull, P. Werner, O. Parcollet, and M. Troyer, Europhys. Lett.
Sémon, G. Sordi, and A.-M. S. Tremblay, Phys. Rev. B 94, 82, 57003 (2008).
155146 (2016). [22] K. Haule and G. Kotliar, Phys. Rev. B 76, 104509 (2007).
[9] H. Terletska, T. Chen, and E. Gull, Phys. Rev. B 95, 115149 [23] J. L. Smith and Q. Si, Phys. Rev. B 61, 5184 (2000).
(2017). [24] E. Gull, P. Staar, S. Fuchs, P. Nukala, M. S. Summers, T.
[10] H. Terletska, T. Chen, J. Paki, and E. Gull, Phys. Rev. B 97, Pruschke, T. Schulthess, T. Maier, Phys. Rev. B. 83, 075122
115117 (2018). (2011).
[11] M. Jiang, U. R. Hahner, T. C. Schulthess, and T. A. Maier, Phys. [25] T. A. Maier, M. S. Jarrell, and D. J. Scalapino, Phys. Rev. Lett.
Rev. B 97, 184507 (2018). 96, 047005 (2006).
[12] J. Paki, H. Terletska, S. Iskakov, and E. Gull, Phys. Rev. B 99, [26] D. J. Scalapino, Chapter 13 in the Handbook of High Temper-
245146 (2019). ature Superconductivity, edited by J. R. Schrieffer (Springer,
[13] M. Schuler, E. G. C. P. van Loon, M. I. Katsnelson, and T. O. Berlin, 2006).
Wehling, SciPost Phys. 6, 067 (2019). [27] M. Jarrell, T. Maier, C. Huscroft, and S. Moukouri, Phys. Rev.
[14] R. Rausch and M. Peschke, New J. Phys. 22, 073051 (2020). B 64, 195130 (2001).

024510-6