Introduction

Heavy fermion (HF) systems in rare-earth (RE) intermetallic compounds originating from hybridization between localized f-electrons and conduction electrons, namely c-f hybridization, are central topics in the field of the strongly-correlated electron systems1. At low temperatures, depending on the strength of the c-f hybridization, the physical properties change from itinerant f electrons because of the Kondo effect to a magnetic order originating with magnetic moments of localized f electrons due to Ruderman–Kittel–Kasuya–Yosida (RKKY) interactions. The competition between itinerant and localized characters of the f-electrons make a quantum critical point (QCP), resulting in the emergence of fertile quantum phenomena such as non-Fermi liquid behavior, and non-BCS HF superconductivity2,3.

On the other hand, the dimensionality in the system characterizes the fundamental physical property. In low-dimensional systems, the enhancement of the electron-electron correlation and/or breaking of the inversion symmetry leads to novel quantum states such as Rashba-type spin-splitting4, Tomonaga–Luttinger liquid5,6, and unconventional superconductivity7,8. The combination of the HF state and low dimensionality modifies the ground state of the system because the order parameter of these systems is much more sensitive to dimensionality9,10. The ground state of two-dimensional (2D) HF can be easily controlled to the vicinity of a quantum critical point, which is the host to realize unconventional physical properties such as HF superconductivity, by simple external fields such as gate-tuning11,12, and surface doping13 in addition to traditional external perturbations; temperature, pressure, and magnetic field. Fabricating artificial low-dimensional strongly correlated electron systems and quantizing a three-dimensional HF state by quantum confinement14 are suitable methods to investigate the novel electronic phase. In the Ce-based artificial superlattice, the suppression of antiferromagnetic (AFM) ordering as well as the increase of the effective electron mass with decreasing of the thickness of the Ce-layer15 and the emergence of the strong-coupling superconductivity16 have been reported. To understand the fundamental properties of 2D HF systems, it is necessary to clarify the electronic band structure and the formation mechanism of the HF. However, the details have remained unclear due to the lack of promising materials and the extremely low transition temperatures of less than a few kelvin to HF even in known materials14,15,17,18.

The growth of the well-ordered atomically thin film on a single crystal substrate is a suitable technique to access such 2D electron systems. So far, various 2D Kondo-lattice has been fabricated on the substrates; multilayer CePt5 thin-film on Pt(111)19, CePb3 on Si(111)20, Graphene on SmB621, and a checkerboard pattern of organic molecules on Au(111)22. However, Yb-based 2D HF material, in which the Yb ion is the most fundamental element to realize HF23,24 and has a symmetrical electronic-hole configuration to the Ce one, has not been reported. In particular, the RE-based monoatomic layer Kondo-lattice showing HF state has never been reported.

In this study, we report the HF electronic structure of a novel Yb-based monoatomic layer Kondo lattice; synchrotron-based angle-resolved photoemission spectroscopy (ARPES) on monoatomic layered YbCu2 on Cu(111). The surface atomic structure of YbCu2 on Cu(111) is shown in Fig. 1a. The Yb atoms surrounded by Cu atoms are arranged in a triangular lattice. In a similar surface alloy RE NM2/NM(111) (NM = noble metal), various physical properties appear such as ferromagnetic ordering25,26,27 and Weyl nodal-line fermion28 depending on the containing RE element, but there is no report on the appearance of HF character so far. Figure 1c, d shows the LEED patterns of the Cu(111) substrate and the Yb-adsorbed Cu(111) surface at 70 K, respectively. In addition to the primitive (1 × 1) spots originating from the Cu(111) substrate indicated by yellow arrows, the (\(\sqrt{3}\times \sqrt{3}\))R30 structure with the Moiré patterns, originating from the small lattice mismatch between Cu(111) and the topmost surface alloy layer, was observed, indicating the successful fabrication of the monoatomic YbCu2 layer, one possible model of the Yb-Cu surface alloy system, on the Cu(111) substrate. Note that the overall trend of the LEED patterns is consistent with those of other RE NM2/NM(111) systems25,26,27,28,29,30.

Fig. 1: Monoatomic-layer YbCu2 on Cu(111) substrate.
figure 1

a A surface atomic structure of YbCu2/Cu(111). b Top view of monoatomic-layer YbCu2. The dashed line indicates the unit cell of YbCu2. c LEED pattern of Cu(111)-(1 × 1) substrate. d Same as (b) but for YbCu2/Cu(111)-(\(\sqrt{3}\times \sqrt{3}\))R30. Both LEED patterns were taken at the temperature of 70 K. The primitive (1 × 1) and (\(\sqrt{3}\times \sqrt{3}\))R30 are indicated by orange and blue arrows, respectively. The distortions of the LEED image are due to the flat microchannel plate used for the LEED measurement. The satellite spots around the integer spots represent the moiré superstructure originating from a small lattice mismatch of YbCu2 and Cu(111).

Results and discussion

The itinerant or localized character of Yb 4f electrons is strongly reflected in the valency of the Yb ions. Figure 2a shows Yb 3d core-level spectrum of YbCu2/Cu(111) at 15 K. The photoelectron peaks at the binding energies of 1528 and 1538 eV originate from the Yb2+ and Yb3+3d final states, respectively, after photoexcitation. From the intensity ratio between the Yb2+ and Yb3+ peaks after subtracting the background indicated by the dotted line in the figure, the mean valence of Yb ions was evaluated as 2.41 ± 0.01. To confirm the consistency of the coexistence of Yb2+ and Yb3+ observed in the Yb 3d core-level spectra to the electronic state near the Fermi level (EF), angle-integrated valence-band photoelectron spectra of the Cu(111) clean substrate and the YbCu2/Cu(111) surface are shown in Fig. 2b. The Cu 3d states at the binding energy of 3 eV are dominant in the Cu(111) substrate. In the YbCu2/Cu(111) spectrum, there are two narrow peaks originating from the Yb2+4f spin-orbit pair near EF, and broad peaks of Yb3+4f final states and Cu 3d states at the binding energy of 3–13 eV. These results strongly suggest that the Yb ions in monoatomic layer YbCu2 are mixed valence. Note that in the YbAu2/Au(111), which has a similar atomic structure to YbCu2/Cu(111), Yb ions are almost divalent29. The reason for the difference in the Yb valence between YbCu2 and YbAu2 would be due to the in-plane lattice compression, which can be explained by the analogy from the bulk Yb-based intermetallic compounds under high pressure31, because the lattice constant of bulk Cu is about 10% smaller than that of Au. The lattice compression would promote the valence transition from Yb2+ to Yb3+ due to the smaller ionic radius of Yb3+ than that of Yb2+, realizing a mixed-valence state in YbCu2/Cu(111).

Fig. 2: Surface electronic structures of YbCu2/Cu(111).
figure 2

a Yb 3d core-level spectrum of YbCu2/Cu(111) taken with 1650-eV photons at the temperature of 15 K. Black circles and red lines represent the raw data and fitted curve, respectively. The dotted line indicates the Shirley-type background. b Angle-integrated valence-band spectra of Cu(111) (black) and YbCu2/Cu(111) (red) taken with horizontally polarized 35-eV photons at 10 K. c A hexagonal surface Brillouin zone of YbCu2/Cu(111). kx and ky are defined along \(\bar{{{\Gamma }}}\)--\(\bar{{{{{{\rm{K}}}}}}}\) and \(\bar{{{\Gamma }}}\)--\(\bar{{{{{{\rm{M}}}}}}}\) of YbCu2. d, e ARPES intensity plots along \(\bar{{{\Gamma }}}\)--\(\bar{{{{{{\rm{K}}}}}}}\) and \(\bar{{{\Gamma }}}\)--\(\bar{{{{{{\rm{M}}}}}}}\) taken with horizontally polarized 37-eV photons at 7 K. ARPES intensities are divided by the Fermi–Dirac distribution function convolved with the instrumental resolution. f Constant energy contours taken with the energy window of ±15 meV using 35-eV photons at 7 K. The contour at the binding energy of 0 eV corresponds to an experimental Fermi surface.

Figure 2d, e shows ARPES band dispersions at 10 K along \(\bar{{{\Gamma }}}\)\(\bar{{{{{{\rm{K}}}}}}}\) and \(\bar{{{\Gamma }}}\)\(\bar{{{{{{\rm{M}}}}}}}\), respectively, in the hexagonal surface Brillouin zone (SBZ) shown in Fig. 2c. The flat band is close to EF and highly dispersive bands are observed near the \(\bar{{{\Gamma }}}\) point. According to the previous study and the DFT calculation for other RE NM2/NM(111) families, the flat band and well-dispersive bands mainly originate from the Yb2+ 4f7/2 and the mixing of the Yb 5d and Cu sp and d orbitals, respectively26,29,32. The detailed assignments of these bands are shown in Supplementary Note 1. It should be noted that the photoelectron intensities of the dispersive band near the \(\bar{{{\Gamma }}}\) point at the positive wavenumber region are relatively weak in both \(\bar{{{\Gamma }}}\)\(\bar{{{{{{\rm{K}}}}}}}\) and \(\bar{{{\Gamma }}}\)\(\bar{{{{{{\rm{M}}}}}}}\) directions due to a photoexcitation selection rule33. The energy position of the Yb2+ 4f7/2 is very close to EF, which is the general feature of the Yb-based HF system such as YbRh2Si223 and β-YbAlB424, suggesting the same mixed valent character of Yb ions in YbCu2/Cu(111) as the result of Yb 3d peaks. It should be noted that the surface electronic structure of almost all HF materials tends to be localized, which is inconsistent with that of the bulk, due to a surface lattice expansion. Also, the Yb-ions in bulk YbCu2 are mixed-valence34,35, which is similar to YbCu2/Cu(111), but the orthorhombic crystal structure is different from that of YbCu2/Cu(111). Additionally, the surface state of bulk YbCu2 is divalent, which is consistent with other Yb compounds. Therefore, the origin of the mixed-valent character of YbCu2/Cu(111) is not the same as that of bulk YbCu2.

In Figure 2d, e, the Yb2+ 4f flat band is modulated at the cross points to the conduction bands just below EF, red providing evidence of c-f hybridization36. It should be noted that the c-f hybridization bands can appear in periodically located Yb and Cu atoms on the surface, not in randomly diluted Yb impurities in bulk Cu. The hole bands at the \(\bar{{{\Gamma }}}\) point and the Yb2+ 4f7/2 states near EF can be confirmed to originate from the YbCu2 layer by calculations26,29,32 (see Supplementary Fig. S2). Figure 2f shows a series of the constant energy contours of the YbCu2/Cu(111) surface. In the map at the binding energy of 0 eV, which corresponds to the experimental Fermi surfaces, there are strong photoelectron intensity areas at \(\bar{{{\Gamma }}}\) point and near the zone boundary. However, as shown in Fig. 2d, e, non-bonding 4f character only appears near the \(\bar{{{{{{\rm{K}}}}}}}\) and \(\bar{{{{{{\rm{M}}}}}}}\) points, but the c-f hybridization feature only exhibits near the \(\bar{{{\Gamma }}}\) point. Therefore, in the following part, we focus on the hybridized band near the \(\bar{{{\Gamma }}}\) point to investigate in detail the HF character appearing in monoatomic layered YbCu2.

In the HF system, the size of the Fermi surface is modulated by the changing of the temperature due to the enhancement of the c-f hybridization. Figure 3a shows the temperature-dependent ARPES images along \(\bar{{{\Gamma }}}\)\(\bar{{{{{{\rm{K}}}}}}}\). The overall trend of those images is consistent except near EF. To reveal the change of the dispersion at EF in more detail, the momentum-distribution curves (MDCs) at EF are plotted in Fig. 3b. The peak position of the hybridized band near the \(\bar{{{\Gamma }}}\) point at 130 K is shifted toward a higher wavenumber at 7 K, suggesting an enlargement of the Fermi surface by the development of the c-f hybridization at low temperature.

Fig. 3: Two-dimensional heavy fermion state in YbCu2/Cu(111).
figure 3

a ARPES intensity plots along \(\bar{{{\Gamma }}}\)--\(\bar{{{{{{\rm{K}}}}}}}\) at 130 and 7 K taken with horizontally polarized 37-eV photons. ARPES intensities are divided by the Fermi–Dirac distribution function convolved with the instrumental resolution. b Momentum distribution curves at EF taken from (a) with the energy windows of ±10 meV. Arrows indicate the peak positions of the MDCs. c Magnified ARPES image near the \(\bar{{{\Gamma }}}\) point taken with circularly polarized 35-eV photons at 15 K. ARPES intensities are divided by the Fermi–Dirac distribution function convolved with the instrumental resolution. The filled and break lines indicate the simulated band dispersions \({{E}_{k}}^{\pm }\) with Vk = 120 meV and 0 meV by the PAM. The open and filled circles indicate the peak positions from energy distribution curves (EDCs) and MDCs, respectively. d Photon-energy dependence of MDCs at the normal emission at binding energies of 0 eV (upper panel) and 250 meV (lower panel) with the energy windows of ±10 meV. Dashed lines indicate the guide of the MDC peak position by eye.

Figure 3c shows the ARPES image around the \(\bar{{{\Gamma }}}\) point taken with circularly polarized photons at 15 K. The c-f hybridization branches S1 and S2 are visible. To determine the dimensionality of the c-f hybridization bands, the photon-energy dependence of ARPES was measured as shown in Fig. 3d. Both S1 and S2 bands show no photon-energy dependence, indicating no out-of-plane (kz) dispersion. These experimental results strongly suggest that the c-f hybridization band is formed in the 2D YbCu2 plane. To evaluate the c-f hybridization feature, comparing it to the periodic Anderson model (PAM) is helpful37. In the case of the Coulomb repulsion energy between 4f electrons Uff is zero or infinity, the band dispersions \({{E}_{k}}^{\pm }\) of PAM is described as

$${{E}_{k}}^{\pm }=\frac{{\epsilon }_{c}+{\epsilon }_{f}\pm \sqrt{{({\epsilon }_{c}-{\epsilon }_{f})}^{2}-4{{V}_{k}}^{2}}}{2}$$
(1)

where ϵc and ϵf are the dispersions of the conduction band and the 4f band, respectively, and Vk is the hybridization intensity. For the fitting using the PAM, a k-linear hole band dispersion was assumed for the conduction band to reproduce the steep band shape observed by ARPES. From the fitting by Equation (1), ϵf and Vk are evaluated as 0.06 and 0.12 eV, respectively. The fitting results are shown in Fig. 3c. The solid and dashed lines indicate the band dispersions \({{E}_{k}}^{\pm }\) with Vk = 0.12 eV and 0 eV, respectively. From the PAM analysis, the Fermi velocity vF and Fermi wavenumber kF of the bare unhybridized conduction band are evaluated as 4.77 eV Å and 0.004 Å−1, respectively. The effective mass of bare conduction band mb becomes 5.65 × 10−33 kg. Note that the shape of the simulated bare conduction band is good agreement with the ARPES image at room temperature as shown in Fig. S7 in Supplementary Note 5. On the other hand, by the fitting of the ARPES band dispersion near EF at 15 K, vF and kF of the 2D HF band are 0.58 eV Å and 0.055 Å−1, respectively, and the effective mass m* of the HF band is 6.56 × 10−31 kg. By comparing the experimentally obtained heavy conduction band at 15 K and the simulated bare conduction band, a mass enhancement factor of the effective mass m* of the HF state at 15 K from the unhybridized one mb (m*/mb) is evaluated as about 120 suggesting the appearance of heavy quasiparticles at low temperatures.

We now discuss the temperature dependence of the quasiparticle peak just below EF, so-called Kondo resonance (KR) peak. In the HF system, the temperature dependences of the energy position and intensity of the KR peak are reflected in the spectral weight transfer between 4f state and the conduction band as well as renormalization due to the development of the c-f hybridization. Figure 4a1 shows the angle-integrated photoelectron intensity near EF as a function of temperature, and Fig. 4a2 is the same, but the intensities are divided by the Fermi–Dirac distribution function convolved with the instrumental resolution. The KR peak energy is shifted to the EF side with decreasing temperature, indicating the evolution of the renormalization due to the HF formation. To discuss the temperature-dependent development of the HF state in more detail, the peak positions and intensities are obtained from the fitting of Fig. 4a2 by a Lorentzian function after subtracting a Shirley-type background, as shown in Fig. 4d, e. The peak position shifts from 42 meV to 22 meV with decreasing temperature and is saturated at 30 K. The integrated intensity increases with decreasing temperature and is also saturated at 30 K. According to photoelectron spectroscopic studies of bulk RE intermetallic compounds, such saturated temperature represents a coherence temperature (Tcoh), at which the c-f hybridization state is fully established, resulting in a HF state38,39. Tcoh sets the temperature scale on which the weight increases and saturation typically occurs below Tcoh. This result is evidently visible in spectral properties like ARPES, while response quantities, such as conductivity and specific heat, are less clear. It has been shown by Bethe ansatz solution of the single-impurity problem and numerous calculations for lattice problems40. The spectral weight is strongly suppressed towards EF, even after the spectra are divided by the Fermi–Dirac distribution function, suggesting that the c-f hybridization gap appears below Tcoh. The above experimental fact strongly suggests that YbCu2 is the 2D monoatomic-layered HF material with Tcoh ~ 30 K, which is more valence fluctuative than other low-dimensional HF such as CeIn3/LaIn3 superlattice (Tcoh = 1.6 K)15.

Fig. 4: Momentum dependence of temperature-evolution of Kondo resonance peak by temperature.
figure 4

a1 Angle-integrated photoelectron spectra near EF as a function of temperature taken with horizontally polarized 35-eV photons. a2 same as (a1) but the intensity is normalized by the Fermi–Dirac distribution function convolved with the instrumental resolution. b The same ARPES image as Fig. 2d indicates the kx positions where the temperature-dependent peak energy and intensity were measured. c Angle-resolved photoelectron spectra near EF as a function of the temperature. The kx positions at 0.5, 0.0, and −0.1 Å−1 are representative of the local f state only, the \(\bar{{{\Gamma }}}\) point, and the crossing point of the c-f hybridization, respectively. d Momentum dependence of the energy position of the quasiparticle peak plotted on a linear scale of temperature. e Momentum dependence of the intensity of the quasiparticle peak plotted on a logarithmic scale of temperature. The shaded area indicates the HF state estimated by the saturated temperature in angle-integrated data, and the boundary between the shaded area and the white background region indicates the experimentally evaluated coherence temperature Tcoh. Error bars are determined by the standard deviation of the fits to the data.

To investigate the momentum-dependent c-f hybridization formation, we took the temperature-dependent peak position and intensity of the quasiparticle peak at three wavenumbers (kx = 0.5, 0.0, −0.1 Å−1) as shown in Fig. 4d, e. The change of the peak position at kx = 0.5 Å−1, which is the cut only the Yb2+ 4f7/2 state, almost follows the angle-integrated one. This suggests that the angle-integrated spectrum mainly focuses on the high-density Yb2+ 4f state and the renormalization is effective for all 4f states. In contrast to the saturated feature in the angle-integrated spectrum at T = 30 K, the KR peak positions at kx = 0.0 and −0.1 Å−1 are shifted toward the higher-binding energy side below Tcoh suggesting the hybridization gap enlargement, even though the peak shifts at kx = 0.0 and −0.1 Å−1 also follows the angle-integrated peak above Tcoh. On the other hand, the peak intensity, which corresponds to the spectral weight transfer between Yb 4f states and conduction bands, continued to increase below Tcoh as shown in Fig. 4e. To the best of our knowledge, similar behavior has not been reported yet except for the angle-integrated photoemission spectroscopy of a Kondo semiconductor SmB641,42, which originated from the transition from the metallic to the semiconducting state, but the magnitude of the peak shift observed here is much larger than that of SmB6. The peak shift in SmB6 is immediately saturated just below Tcoh, which is not consistent with the behavior in YbCu2, suggesting a different mechanism of the KR peak shift. Both the peak shift and the developing intensity below Tcoh of YbCu2 suggest that the 2D HF state still develops even below Tcoh. Further theoretical analysis such as the dynamical mean field theory about the development of the HF state in the 2D system would help the understanding of the temperature-dependent behavior of the KR peak.

In conclusion, we report the electronic structures of a novel Yb-based monoatomic layered Kondo lattice YbCu2 on Cu(111) by ARPES. Our spectroscopic data provide direct evidence of the appearance of a purely 2D HF state with Tcoh = 30 K, which is extremely higher than other 2D HF materials15,17,18, in a monoatomic layered material for the first time. Monoatomic layered YbCu2 is the minimal material to realize low-dimensional HF containing RE elements and act as a building block to reveal novel electron-correlation-driven phenomena, for example, the proximity effect of layered material between 2D HFs and other many-body interactions such as superconductivity and magnetism. Quantum fluctuations in 2D materials are much more sensitive to external fields18. The ground state of YbCu2 would be tuned around QCP by external perturbations conventionally applied to other 2D materials, where surface carrier doping by alkali metal adsorption and gate-tuning of carrier concentration by the growth of Cu(111) ultrathin film on an insulating substrate such as sapphire may be realized. Such novel techniques are expected to explore novel quantum critical phenomena in 2D materials, such as atomic-layer unconventional superconductivity.

Methods

Sample preparation

Cu(111) substrate was cleaned by Ar-ion sputtering with an acceleration energy of 0.5 keV and annealing at 800 K. After several sputtering and annealing cycles, a sharp (1 × 1) low-energy electron diffraction (LEED) pattern was confirmed as shown in Fig. 1c. Yb atoms were evaporated on the Cu(111) substrate at 600 K. Because the crystallinity of the YbCu2 layer was quite sensitive to the substrate condition such as cleanness of surface and growth temperature, we precisely monitored the sharpness of the diffraction from YbCu2 and moiré pattern, which is directly linked to the crystal quality of the YbCu2/Cu(111), by the reflective high energy electron diffraction (RHEED) in the growth process.

Photoemission experiments

ARPES and core-level photoemission spectroscopy measurements were performed at BL-2A MUSASHI of the Photon Factory, and BL7U SAMRAI43 of the UVSOR-III Synchrotron Facility. The energy resolution and the energy position of the Fermi-level were calibrated by the Fermi-edge of polycrystalline Au films electrically contacted to the sample holder. Energy resolutions for ARPES and core-level photoemission spectroscopy were better than 20 meV and 100 meV, respectively. In temperature-dependent measurements, the position of EF and the instrumental resolution were accurately calibrated by measuring the Fermi edge of the Au thin film at all measurement temperatures.

Band calculations

Band structures of freestanding YbCu2 and YbCu2/Cu(111) slab were calculated by using the WIEN2K code44 including spin-orbit interaction within the generalized gradient approximation of the Perdew, Burke, and Ernzerhof exchange-correlation potential45. The in-plane lattice constant of the YbCu2 was set to the experimentally obtained value (4.80 Å) from LEED measurements. The atomic structure of YbCu2/Cu(111) was modeled by a symmetric slab of six layers of Cu with a surface covered with YbCu2 layers. No electron correlation was included, in the band calculations. To obtain the overall trend of the electronic structure, such a condition that does not include the electron correlation would also be sufficient. The calculated band structures are shown in Figs. S1 and S2.