betaFeedback

WorksPhysics and Astronomy

Local excitation of kagome spin ice magnetism in HoAgGe seen by scanning tunneling microscopy

Abstract

The kagome spin ice can host frustrated magnetic excitations by flipping its local spin. Under an inelastic tunneling condition, the tip in a scanning tunneling microscope can flip the local spin, and we apply this technique to kagome metal HoAgGe with a long-range ordered spin ice ground state. Away from defects, we discover a pair of pronounced dips in the local tunneling spectrum at symmetrical bias voltages with negative intensity values, serving as a striking inelastic tunneling signal. This signal disappears above the spin ice formation temperature and has a dependence on the magnetic fields, demonstrating its intimate relation with the spin ice magnetism. We provide a two-level spin-flip model to explain the tunneling dips considering the spin ice magnetism under spin-orbit coupling. Our results uncover a local emergent excitation of spin ice magnetism in a kagome metal, suggesting that local electrical field induced spin flip climbs over a barrier caused by spin-orbital locking.

A kagome lattice is made of corner-sharing triangles. Based on its special geometry, the kagome lattice can host frustrated magnetic or electronic structures, including quantum spin liquids [1], spin ices [2, 3] and flat electronic bands [4, 5]. Recent years have witnessed substantial advances in exploring the emergent physics of metallic kagome quantum materials using electronic structure methods [4, 5], highlighting the interplay between geometry, topology, correlation, and spin. However, the quantum spin liquid and spin ice phase in the kagome lattice are largely unexplored by advanced electronic structure methods. In this work, we advance this research frontier by detecting pronounced tunneling conductance dips in kagome spin ice HoAgGe, serving as the signature of local spin-flip that is crucial for harnessing frustrated spin ice excitations.

In spin ice materials, the local magnetic spins respect “ice rules” similar to the electrical dipoles in water ice [2, 4, 6, 7, 8, 9, 10, 11, 12, 13]. The spin ice is typically found in quantum materials with three-dimensional pyrochlore lattices, and has been recently discovered in a kagome metal HoAgGe [13]. In a kagome spin ice, the ice rule requires 2-in-1-out or 2-out-1-in spin configurations in each triangle [Fig. 1(a)]. Upon lowering the temperature, it is believed that spins first form a short-range ice order and then a long-range ice order. Intriguingly, when locally breaking the ice rule by flipping a certain spin, spin configurations with three-in or three-out can emerge, in analogy to the magnetic monopoles predicted by Dirac [24]. Creating spin excitations in a controlled and local manner can open new avenues for spin ice research.

HoAgGe consists of distorted Ho3+^{3+} kagome lattices with layer stacking [Fig. 1(b)], where the strong local easy-axis anisotropy together with the ferromagnetic nearest-neighbor coupling of the Ho3+^{3+} moments lead to “1-in-2-out” or “2-in-1-out” ice rules on the kagome lattice [Fig. 1(c)] [13]. The atomic structure of HoAgGe single crystal [Fig. 1(d)] is demonstrated by the high angle annular dark-field-scanning transmission electron microscopy in an aberration-corrected transmission electron microscopy, as shown in Figure 1(e). The high-resolution image and corresponding elemental maps clearly show the Ho kagome structure in the aa-bb plane with the well-identified Ho, Ag and Ge atoms. Its short-range spin ice order occurs at around 20K, while its long-range spin ice order occurs at 11.6K [13]. In the spin ice state, magnetization versus external magnetic fields parallel to the kagome plane exhibits a series of plateaus [Fig. 1(f)], and the corresponding field-induced phases detected by neutron scattering are all consistent with the spin ice rule [13]. Magnetization along the cc-axis, on the other hand, shows a gradual increment with field strength [Fig. 1(f)] that is consistent with a ferromagnetic canting of the spins toward the cc-axis [13]. The temperature-dependent resistivity data [Fig. 1(g)] shows a peak feature near the spin ice transition and demonstrates metallic behavior. Consistently, our angle-resolved photoemission spectroscopy characterization of this material at the spin ice state (TT = 8K) reveals both Fermi surfaces and multiple dispersive bands at the Fermi level [Fig. 1(h)]. The magnetization plateaus and dispersive bands together demonstrate the coexistence of local moments and itinerant electrons. This situation allows us to use the electronic tunneling method to introduce spin excitations that we further explore.

Coexistence of local moment and itinerant electrons in kagome spin ice HoAgGe

Figure 1. Coexistence of local moment and itinerant electrons in kagome spin ice HoAgGe. (a) Schematic for a kagome spin ice. (b) Crystal structure of HoAgGe showing layered structure. (c) The distorted Ho kagome lattice with spins following the ice rule, the shaded lines show a perfect kagome lattice. (d) Image of a HoAgGe single crystal with its cc-axis along the horizontal direction. (e) High-angle annular dark-field-scanning transmission electron microscopy image (left) and corresponding energy-dispersive spectrometry elemental maps (right), imaged along cc-axis. (f) Low-temperature magnetization curves measured at 2K, where the in-plane magnetization curve shows fractional field-induced transitions. (g) Resistivity curve showing metallic behavior of HoAgGe. (h) Fermi surfaces (left) and band dispersions along the high symmetry lines (right) detected by angle-resolved photoemission at 8K.

Compared to other techniques, scanning tunneling microscope can flip the spin of a single atom through inelastic tunneling [25, 26, 27, 28, 29, 30, 31], and is highly desirable to be applied to study kagome spin ice [Fig. 2(a)]. Above a threshold tunneling bias voltage, electrons can transfer energy to the spin-flip excitations. In this work, we explore this frontier in kagome spin ice HoAgGe. Owing to the strong interlayer bonding, it is extremely rare to obtain a flat atomic surface by cryogenic cleaving, which severely challenges its scanning tunneling microscopy study. Therefore, we have to cleave more than 50 crystals and scan extensively for each cleaved crystal to find the atomic flat surface with unit-cell step heights [Figs. 2(b) and (c)]. The surface exhibits a hexagonal lattice with its lattice constant consistent with that of the crystal [Fig. 2(d)]. In the crystal structure, only Ge atoms in the HoGe layer form a hexagonal lattice. Thus, we identify this surface as HoGe surface termination. At low-temperature TT = 0.3K, we measure the differential conductance spectrum in the clean lattice region and detect a pair of pronounced dips located at Ed=±107E_d = \pm107 meV [Fig. 2(e)], with corresponding kinks in the tunneling current data Figure 2(f). Such dips are the most pronounced features in the data and extra tiny spectral dips (not reaching to negative value) and bumps are currently not well understood. The energy locations of this pair of dips are symmetric, and their corresponding differential conductance values are strongly negative. These two features indicate the dips are from inelastic tunneling rather than from the density of states of the band structure, as also confirmed by the absence of such a feature around −0.1-0.1eV in our photoemission data in Figure 1(h). These titanic spectral features are different from the mild step features for inelastic tunneling of spin-flip of an isolated magnetic atom [25, 26, 27, 28, 29]. To understand their origin, we systematically examine their spatial, temperature, and magnetic responses in Figure 3.

Schematic, topographic images, height profile, and tunneling spectra of kagome spin ice HoAgGe

Figure 2. Pronounced tunneling dips in kagome spin ice HoAgGe. (a) Schematic for scanning tip probe induced local spin-flip in a kagome spin ice. (b) Topographic image for a large field of view showing atomic flat cleaving surfaces with unit-cell height steps. (c) Height profile along the red line drawn in (b), showing atomic step height of 4.1 Å that is consistent with the unit cell height. (d) A zoomed-in image of the topographic image (up) whose lattice symmetry is consistent with the HoGe atomic layer (down). (e) Differential conductance spectrum on the clean region of HoGe surface showing a pair of pronounced electronic dips occurring at Ed=±107 mVE_d = \pm107\ \mathrm{mV}. (f) The corresponding kinks in the tunneling current spectrum.

Spatial, temperature, and magnetic field dependence of the tunneling dips

Figure 3. Spatial, temperature, and magnetic field dependence of the tunneling dips. (a) Topographic image of a large cleavage surface. The surface contains adatoms and vacancies as impurities caused by cleavage. (b) Differential conductance map g(r,E)g(r, E) at the dip energy, showing inhomogeneity. (c) Anticorrelation between the dips and impurities. We find the dips are generally away from the impurities, indicating that dips are intrinsic features of the clean lattice. (d) Intensity plot of a series of differential conductance spectra along the red line drawn in (a) and (b), showing the spatial inhomogeneity of dips at Ed=±107E_d = \pm107 meV. (e) Temperature dependence of the dips at a fixed position, showing its disappearance above 20 K. The inset shows the differential heat capacity reproduced from Ref. 13, where the short-range spin ice formation temperature is estimated to be around 20 K. (f) Magnetic field dependence of the dip at a fixed position. An in-plane magnetization (the sample riches the 1/3 magnetization plateau with B=1.5B = 1.5 T applied along the bb-axis) enhances the depth of the dip, while an out-of-plane magnetization (the sample riches the magnetization plateau with B=6B = 6 T applied along the cc-axis) suppresses the depth of the dip. The dash lines mark the respective zero-value.

We perform a spectroscopic map at the dip energy for a large area in Figure 3(a) and (Figure 3b). The spectroscopic map data is inhomogeneous, and the low-intensity parts of the map correspond to the emergence of dips, as further supported by representative differential conductance spectra along a line cut [Fig. 3(d)]. Owing to the strong interlayer bonding, the cleavage leaves vacancies and adatoms as surface defects, and we find that the dips are generically away from these defects as illustrated in Figure 3(c), where we mark the positions of both the impurities and dips. Hence the local dip feature can be an intrinsic feature of the clean lattice rather than from impurities, and the inhomogeneity is largely caused by cleavage-induced surface defects that have a long-range suppression of the dip feature. When we raise the temperature, we find that while the dip energy does not change, the dip becomes progressively shallow, and eventually disappears above 20K [Fig. 3(e)]. It is close to the short-range spin ice order transition temperature, as evidenced by the differential heat capacity [reproduced in the inset of Figure 3(e)]. At the defect-free region, we further find that the external magnetic field also does not change dip energy [Fig. 3(f)], but only affect its depth. An in-plane magnetization (the sample riches the 1/3 magnetization plateau with B=1.5B = 1.5T applied along the bb-axis [13, 32]) enhances the depth of the dip, while an out-of-plane magnetization (the sample riches the magnetization plateau with B=6B = 6T applied along the cc-axis) suppresses the depth of the dip. We note that after applying the magnetic field or varying the temperature, we relax the system and relocate our tip to the same atomic position to take the differential spectrum. We clarify that these data are representative data rather than statistical data. The spatial, temperature, and magnetic field responses thus support the intimate relation between the pronounced tunneling dip and spin ice magnetism of the kagome lattice, and we discuss its possible microscopic origin.

In terms of magnetic excitations, the common inelastic tunneling signal can result from the spin-flip process [25, 26, 27, 28, 29]. The spin-flip energy ΔS\Delta_{\mathrm{S}} is related to the magnetic exchange interaction on the order of 11–1010meV, inferred from the short-range spin ice transition temperature of 20K. As Ho carries heavy 4f4f orbitals with a large spin-orbit coupling, we need to consider the orbital flip locked to the spin-flip. Since there is no apparent orbital ordering from transport, the final orbital flipped state would not alter the dip energy too much; however, the intermediate orthogonal orbital configuration can be at much higher energy in the anisotropic crystal field, causing an energy barrier ΔO\Delta_{\mathrm{O}}. Therefore, the spin-flip process here can be described by a two-level system with a small spin energy difference of ΔS\Delta_{\mathrm{S}} and a large orbital barrier ΔO\Delta_{\mathrm{O}} [Fig. 4(a)]. Notably, a similar vibrationally induced two-level picture has been pointed out to explain spikes in differential conductance in molecular junctions [33]. This model further assumes different tunneling conductance values for the two energy levels, and it can naturally produce a pair of tunneling dips at symmetrical energies, as illustrated by our simulations. [Fig. 4(b)], which strongly supports our two-level interpretation of the data.

To demonstrate the two-level spin-flip scenario, we present quantitative calculations of the two essential energy levels, i.e., ΔS\Delta_{\mathrm{S}} and ΔO\Delta_{\mathrm{O}}. Based on the estimation [13] of local exchange interactions JJ from the magnetization data, we calculate the lowest spin-flip energy ΔS=2J2S2=2 meV\Delta_{\mathrm{S}}=2J_{2}S^{2}=2\ \mathrm{meV} as shown in Figure 4(c), where J2=0.23 meVJ_{2}=0.23\ \mathrm{meV} and S=2S=2. The calculated ΔS\Delta_{\mathrm{S}} aligns well, in terms of the order of magnitude, with the extracted values presented in Figure 4(e). To estimate the energy barrier of spin-flip ΔO\Delta_{\mathrm{O}}, we calculate the crystal electric field (CEF) of Ho3+\mathrm{Ho}^{3+} ion on the cleavage HoGe surface, where the surrounding ligands are Ge4−\mathrm{Ge}^{4-} and Ho3+\mathrm{Ho}^{3+} ions (note that it differs from the bulk CEF). The ∣J=8,Jz⟩\lvert J=8,J_z\rangle multiplet (S=2S=2, L=6L=6) of the Ho3+\mathrm{Ho}^{3+} ion, under the C2vC_{2v} site symmetry of Ho3+\mathrm{Ho}^{3+} ion, splits into 17 singlets [34]. We find that the ground state of Ho3+\mathrm{Ho}^{3+} ion manifests Ising anisotropy, with its local axis lying in the abab plane, aligning with the spin ice state in HoAgGe. Moreover, we note that CEF excitations would be accompanied by spin-flip. In particular, the highest CEF level, with its magnetic easy axis perpendicular to that of the ground state, contributes to the energy barrier during the spin-flip process. This barrier is estimated to be ΔO=102 meV\Delta_{\mathrm{O}}=102\ \mathrm{meV} [Fig. 4(d)], yielding good quantitative agreement with the observed dips at Ed=±107 meVE_d=\pm107\ \mathrm{meV}.

Two-level spin-flip model for tunneling dips

Figure 4. Two-level spin-flip model for tunneling dips. (a) Schematic for two-level spin-flip process. In this model, Ψ0\Psi_0 is the kagome spin ice ground state with a differential conductance σ0\sigma_0, Ψ1\Psi_1 corresponds to the spin-flipped state with a differential conductance σ1\sigma_1, and ΔS=1∼10 meV\Delta_S = 1\sim10\,\mathrm{meV} is their spin energy difference. Ψ2\Psi_2 is the intermediate state with orthogonal spin and orbital configurations, and it has a large energy difference ΔO∼0.1 eV\Delta_O \sim0.1\,\mathrm{eV} from the anisotropic crystal electrical field. The three inset figures illustrate three different states by flipping a local spin in the distorted kagome lattice. (b) A typical simulation result of the two-level spin-flip model showing a pair of tunneling dips. (c) Illustration of three superexchange interactions J1,2,3J_{1,2,3} in the distorted kagome lattice. The minimum spin flip energy is calculated to be 2 meV2\,\mathrm{meV} based on the estimated JJ values from the magnetization data[13]. (d) Calculated orbital barrier energy ΔO=102 meV\Delta_O = 102\,\mathrm{meV} from the largest crystal electric field energy level. (e) Fitting results of the magnetic field dependent tunneling dips by the two-level spin-flip model. The experimental data are symmetrized with respect to zero-energy.

In addition, as illustrated in Figure 4(e), the two-level spin-flip model well reproduces the dependence of tunneling dips on the external magnetic field. Notably, the external field mainly changes the spin-flip energy ΔS\Delta_{\mathrm{S}}. The 1.5T in-plane field alone would not break the in-plane ice rule (at the 1/3 magnetization plateau), but can assist spin-flip of inelastic tunneling for half of the spins in the system by introducing Zeeman energy. Thus, it reduces the minimum spin-flip energy ΔS\Delta_{\mathrm{S}} of the total spin system. As for the out-of-plane magnetic field, it will polarize spins to tilt towards the cc-axis. Consequently, the ΔS\Delta_{\mathrm{S}} now has an additional energy contribution from the nearest magnetic exchange J1J_{1}, leading to a moderate increase in its value.

In summary, we extend the study of kagome materials with electronic structure methods to a kagome spin ice material, and detect pronounced tunneling dips as the signature of local spin flip. We show that the local spin-flip physics in a magnetic lattice through inelastic tunneling is different from the case of a single magnetic atom, and is well described by a two-level model involving a crystal field barrier. Our interpretation of the data leads us to conclude that effective magnetic monopoles can be pairwise created via inelastic tunneling using a scanning tunneling microscope tip. It is also meaningful to further explore the pronounced dip signal with a spin-polarized tunneling tip [35], since the spin-polarized tunneling has been achieved in other kagome materials [36, 37, 38]. Therefore, our exploration of the local spin-flip process in a spin ice system provides an initial step towards harnessing its unusual spin excitations including the magnetic monopoles. The gigantic tunneling signal in a kagome spin ice with classical spins opens up new research opportunities that encourage us to further explore the kagome spin liquid with electronic structure tools.

Acknowledgement

We thank the insightful discussions with Kan Zhao, Xi Dai, Yifeng Yang, and Guangming Zhang. We thank R. J. Cava for sharing his experience of using “titanic” in describing the magnetoresistance in WTe₂. We acknowledge the support from the National Key R&D Program of China (No. 2023YFA1407300, 2023YFF0718403, and 2022YFA1403700), National Science Foundation of China (No. 12374060, 11804402, 12004123 and 12334002), Shenzhen Fundamental Research Program (No. JCYJ20220818100405013 and JCYJ20230807093204010), the Outstanding Talents Training Fund in Shenzhen (No. 202108), Guangdong Basic and Applied Basic Research Foundation (No. 2024A1515030118), Innovative Team of General Higher Educational Institutes in Guangdong Province (No. 2020KCXTD001), Shenzhen Science and Technology Program (No. 20231117091158001), and Guangdong Provincial Quantum Science Strategic Initiative (GDZX2201001). TN acknowledges support from the Swiss National Science Foundation (Project 200021E_198011) as part of the FOR 5249 (QUAST) led by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation).

Supplementary information

Methods

Single crystal growth, transport, and magnetization measurement

Single crystals of HoAgGe were grown by the self-flux method. High purity elemental of Ho pieces (99.99%), Ag grain (99.999%), and Ge powder (99.999%) were weighed by stoichiometric ratio Ho:Ag:Ge = 6:705:235. The mixture was transferred into an alumina crucible and sealed in a vacuumed quartz tube. Then the tube was heated up to 1150°C for 30 h, followed by a slow cooling down to 850°C within 300 h. Single crystals of HoAgGe with hexagonal prism type were separated by centrifugation. The transport measurements were performed on a physical property measurement system (Quantum Design PPMS-14) and a variable temperature system equipped with a lock-in amplifier (NF Corp. LI5650) using the standard four-probe and alternating current transport methods. The AC current (IacI_{\mathrm{ac}}) with frequency f=23f = 23 Hz was applied along a-axis and c-axis respectively.

Magnetic measurements were performed in a superconducting quantum interference device magnetometer (Quantum Design MPMS-7).

Transmission electron microscopy

The thin lamella of HoAgGe single-crystal was prepared by focused ion beam along ⟨0001⟩\langle0001\rangle zone axis with a final 5 kV beam to minimize the sample damage. The microstructure of the prepared sample was investigated by high angle annular dark-field-scanning transmission electron microscopy (HAADF-STEM) in double Cs-corrected TEM (FEI Titan Themis G2) operated at 300 kV, with a super-X energy-dispersive spectrometry (EDS) system. Atomic-scale STEM-EDS mappings were generated in Velox (v3.90) with the radial wiener filter to enhance the feature of elemental maps.

Scanning tunneling microscopy

Single crystals with sizes up to 1mm×1mm×1mm1\mathrm{mm} \times1\mathrm{mm} \times1\mathrm{mm} were cleaved mechanically in situ at 80K in ultrahigh vacuum conditions, and then immediately inserted into the microscope head, already at He4 base temperature (4.2K). We then further cool the microscope head to 0.3K via a He3-based single-shot refrigerator. The magnetic field was applied with a small ramping speed of 1T per 20mins. After ramping the field to a desired value, the superconducting magnet is set in the persistent mode, after which we wait for 1~2h for the system to relax and then find the same atomic position and start to take spectroscopic measurements. Tunneling conductance spectra were obtained with Ir/Pt tips using standard lock-in amplifier techniques with a root mean square oscillation voltage of Vm=0.3meVV_\mathrm{m} = 0.3\mathrm{meV} under applied bias voltage of V=200mVV = 200\mathrm{mV} and tunneling current I=1nAI = 1\mathrm{nA}. Topographic images were taken with tunneling junction set up: V=500V = 500~200mV200\mathrm{mV} I=0.05I = 0.05~0.5nA0.5\mathrm{nA}. The conductance maps and gap map were obtained by taking a spectrum at each location (off feedback loop) with tunneling junction set up: V=200mVV = 200\mathrm{mV}, I=1nAI = 1\mathrm{nA}, and modulation voltage Vm=1mVV_\mathrm{m}=1\mathrm{mV}.

Angle-resolved photoemission spectroscopy (ARPES)

Clean surfaces for ARPES measurements were obtained by cleaving HoAgGe samples in situ in a vacuum greater than 5×10−115 \times10^{-11} Torr. ARPES measurements were performed on the “Dreamline” beamline of the Shanghai Synchrotron Radiation Facility, with an overall energy resolution of 20 meV, angular resolution of 0.1∘0.1^\circ. The data is taken at T=8T = 8 K, with photon energy hν=80h\nu= 80 eV.

Two-level spin-flip model

We have devised a theoretical model for a two-level system, exemplified by a double-well potential structure as delineated in Figure 4(a). The system possesses two distinct spin ground states, Ψ0\Psi_0 and Ψ1\Psi_1, with an energy difference of ΔS\Delta_\mathrm{S} separating them. An excited state, Ψ2\Psi_2, with a higher energy level exists within the system, and it exhibits an energy difference of ΔO\Delta_\mathrm{O} when compared to Ψ0\Psi_0. The tunneling conductance for electrons transitioning from the different spin ground states to the excited state are denoted as σ0\sigma_0 and σ1\sigma_1, respectively. The tunneling current is formulated as

I(V)=(n0σ0+n1σ1+neσ0+σ12)V(1)I(V) = \left(n_0\sigma_0 + n_1\sigma_1 + n_e\frac{\sigma_0+\sigma_1}{2}\right)V \tag*{(1)}

n0n_0, n1n_1, and nen_e represent the electron occupation numbers for different states. Their interrelation is given by the expression for their interrelation is Eq. (2).

dn0dt=−We0+W0e=0dn1dt=−We1+W1e=0dnedt=−W0e+We0−W1e+We1=0n0+n1+ne=1(2)\begin{aligned} \frac{dn_0}{dt} &= -W_{e0}+W_{0e}=0 \\ \frac{dn_1}{dt} &= -W_{e1}+W_{1e}=0 \\ \frac{dn_e}{dt} &= -W_{0e}+W_{e0}-W_{1e}+W_{e1}=0 \\ n_0+n_1+n_e &= 1 \tag*{(2)} \end{aligned}

WW denotes the transition rates between distinct energy eigenstates, which are quantitatively expressed via Fermi’s golden rule.

We0=n0∫−∞+∞f(ξ,eV)[1−f(ξ−Δs,eV)] dE(3)W_{e0}=n_0\int_{-\infty}^{+\infty} f(\xi,eV)[1-f(\xi-\Delta_s,eV)]\,dE \tag*{(3)}

Under the low-temperature limit, we derive the solution for Eq. (3):

∫−∞+∞f(ξ,eV)[1−f(ξ+E,eV)] dE={E+14E−e∣V∣e(E−e∣V∣)/kBT−1⋯⋯ΔE>014−E−e∣V∣e(−E−e∣V∣)/kBT−1⋯⋯ΔE<0(4)\int_{-\infty}^{+\infty} f(\xi,eV)[1-f(\xi+E,eV)]\,dE = \begin{cases} E+\dfrac{1}{4}\dfrac{E-e|V|}{e^{(E-e|V|)/k_BT}-1} \quad\cdots\cdots\Delta E>0 \\ \dfrac{1}{4}\dfrac{-E-e|V|}{e^{(-E-e|V|)/k_BT}-1} \quad\cdots\cdots\Delta E<0 \end{cases} \tag*{(4)}

By solving Eq. (2), (3), and (4), we can infer that:

dn0dt=−We0+W0e=0=−n0ε0+neε1=0dn1dt=−We1+W1e=0=−n0ε2+neε3=0ne=1−n0−n1\begin{aligned} \frac{dn_0}{dt} &= -W_{e0}+W_{0e}=0=-n_0\varepsilon_0+n_e\varepsilon_1=0 \\ \frac{dn_1}{dt} &= -W_{e1}+W_{1e}=0=-n_0\varepsilon_2+n_e\varepsilon_3=0 \\ n_e &= 1-n_0-n_1 \end{aligned}
ε0=14Δs−e∣V∣e(Δs−e∣V∣)/kBT−1ε1=Δs+14Δs−e∣V∣e(Δs−e∣V∣)/kBT−1ε2=14(Δs−Δo)−e∣V∣e[(Δs−Δo)−e∣V∣]/kBT−1ε3=(Δs−Δo)+14(Δs−Δo)−e∣V∣e[(Δs−Δo)−e∣V∣]/kBT−1(5)\begin{aligned} \varepsilon_0 &= \frac{1}{4}\frac{\Delta_s-e|V|}{e^{(\Delta_s-e|V|)/k_BT}-1} \\ \varepsilon_1 &= \Delta_s+\frac{1}{4}\frac{\Delta_s-e|V|}{e^{(\Delta_s-e|V|)/k_BT}-1} \\ \varepsilon_2 &= \frac{1}{4}\frac{(\Delta_s-\Delta_o)-e|V|}{e^{[(\Delta_s-\Delta_o)-e|V|]/k_BT}-1} \\ \varepsilon_3 &= (\Delta_s-\Delta_o)+\frac{1}{4}\frac{(\Delta_s-\Delta_o)-e|V|}{e^{[(\Delta_s-\Delta_o)-e|V|]/k_BT}-1} \tag*{(5)} \end{aligned}

The formula for the occupation numbers is derived through the resolution of Equation 5:

n0=ε1ε2(ε0ε2+ε0ε3+ε1ε2)n1=ε0ε3(ε0ε2+ε0ε3+ε1ε2)ne=ε0ε2(ε0ε2+ε0ε3+ε1ε2)(6)\begin{aligned} n_0 &= \frac{\varepsilon_1\varepsilon_2}{(\varepsilon_0\varepsilon_2+\varepsilon_0\varepsilon_3+\varepsilon_1\varepsilon_2)} \\ n_1 &= \frac{\varepsilon_0\varepsilon_3}{(\varepsilon_0\varepsilon_2+\varepsilon_0\varepsilon_3+\varepsilon_1\varepsilon_2)} \\ n_e &= \frac{\varepsilon_0\varepsilon_2}{(\varepsilon_0\varepsilon_2+\varepsilon_0\varepsilon_3+\varepsilon_1\varepsilon_2)} \tag*{(6)} \end{aligned}

Therefore, derived from the tunneling current Eq. (1), the differential conductance is expressed as Eq. (7).

dIdV=(σ0+σ1)2+(σ0−σ1)2(n0+Vdn0dV)+(σ1−σ0)2(n1+Vdn1dV)(7)\frac{dI}{dV}=\frac{(\sigma_0+\sigma_1)}{2}+\frac{(\sigma_0-\sigma_1)}{2}\left(n_0+V\frac{dn_0}{dV}\right)+\frac{(\sigma_1-\sigma_0)}{2}\left(n_1+V\frac{dn_1}{dV}\right) \tag*{(7)}

The simulated differential conductance is illustrated in Figure 4(e).

Crystal electric field (CEF) calculation

We carry out GGA calculations [39, 40] to investigate the valence states in HoAgGe. Figure S1 presents the orbitally resolved density of states (DOS). The fully occupied Ag 4d4d states are located at 4-6 eV below the Fermi level, suggesting the formal Ag+^{+} valence state. Similarly, the Ge 4p4p states, also fully occupied, lie at 0-4 eV below the Fermi level, of which the large bandwidth is mainly due to band hybridization with both Ag 4d4d and Ho 4f4f states. Consequently, the Ho3+^{3+} valence state is deduced, collaborated by the partially occupied 4f4f states crossing the Fermi level.

The Ho3+^{3+} ion, with its 4f104f^{10} configuration, has a total spin of 2 and a total angular momentum of 6 according to Hund’s rules. Thus, the multiplet with the total momentum J=L+S=8J = L + S = 8 has the lowest energy, specifically the multiplet ∣J=8,Jz⟩\lvert J = 8, J_z\rangle for Jz=−8,−7,…,8J_z = -8,-7,\ldots,8. Within the HoGe surface, the Ho3+^{3+} ion is subject to a CEF originating from the surrounding ligand ions Ge4−^{4-} and Ho3+^{3+}. Given the site symmetry C2vC_{2v} of the Ho3+^{3+} ion, the ∣J=8,Jz⟩\lvert J = 8, J_z\rangle multiplet would be split into 17 singlets. The CEF Hamiltonian can be written as [41],

HCEF=∑k,qBqkOqkH_{CEF} = \sum_{k,q} B_q^k O_q^k

where the BqkB_q^k are CEF parameters and OqkO_q^k are spherical tensor operators. The C2vC_{2v} site symmetry restricts the q=2,4,6q = 2,4,6 and k=0,2,4,6k = 0,2,4,6. To estimate the CEF effect, we adopt a point-charge model via the PyCrystalField [42], setting the ligand distance range to 6 Å. The calculated CEF parameters are obtained as listed in Table 1. Moreover, the CEF eigenstates, expressed in terms of the basis ∣Jz⟩\lvert J_z\rangle, are tabulated in Table 2.

B20B_{2}^{0}B22B_{2}^{2}B40B_{4}^{0}B42B_{4}^{2}B44B_{4}^{4}B60B_{6}^{0}B62B_{6}^{2}B64B_{6}^{4}B66B_{6}^{6}
1.147E-1-7.945E-12.970E-56.275E-4-9.588E-41.400E-7-2.240E-62.430E-6-4.210E-6

Table 1. CEF parameters BkqB_{k}^{q} (in meV) of the Ho ion within the HoGe surface of HoAgGe.

EE (meV)∣−8⟩|-8\rangle∣−7⟩|-7\rangle∣−6⟩|-6\rangle∣−5⟩|-5\rangle∣−4⟩|-4\rangle∣−3⟩|-3\rangle∣−2⟩|-2\rangle∣−1⟩|-1\rangle∣0⟩|0\rangle∣1⟩|1\rangle∣2⟩|2\rangle∣3⟩|3\rangle∣4⟩|4\rangle∣5⟩|5\rangle∣6⟩|6\rangle∣7⟩|7\rangle∣8⟩|8\rangleeasy axis
0.0000.00.03450.00.16460.00.38620.00.5680.00.5680.00.38620.00.16460.00.03450.0xx
0.0000.00960.00.08470.00.26980.00.49290.00.5950.00.49290.00.26980.00.08470.00.0096xx
27.365-0.04210.0-0.25340.0-0.49760.0-0.43180.00.00.00.43180.00.49760.00.25340.00.0421xx
27.3650.00.12610.00.39350.00.51860.00.24550.0-0.24550.0-0.51860.0-0.39350.0-0.12610.0xx
45.917-0.12050.0-0.45830.0-0.43530.00.07820.00.39950.00.07820.0-0.43530.0-0.45830.0-0.1205xx
45.9180.0-0.29240.0-0.52460.0-0.2070.00.31060.00.31060.0-0.2070.0-0.52460.0-0.29240.0xx
60.2680.00.48640.00.35410.0-0.26260.0-0.26270.00.26270.00.26260.00.35410.00.48640.0zz
60.280-0.26290.0-0.53740.0-0.03010.00.37570.00.00.0-0.37570.00.03010.00.53740.00.2629zz
70.3790.47820.00.31950.0-0.33470.0-0.0750.00.32140.0-0.0750.0-0.33470.00.31950.00.4782zz
70.4490.00.58610.0-0.07380.0-0.32780.00.20870.00.20870.0-0.32780.0-0.07380.00.58610.0zz
76.7480.62370.0-0.12760.0-0.19460.00.23830.00.00.00.23830.0-0.19460.0-0.12760.00.6237zz
78.9290.0-0.49130.00.41790.0-0.06380.0-0.28260.00.28260.00.06380.0-0.41790.00.49130.0zz
81.1090.50510.00.040260.0-0.1840.0-0.11810.00.26430.0-0.11810.0-0.1840.00.040260.00.5051yy
88.4830.00.26420.0-0.43850.00.44780.0-0.19350.0-0.19350.00.44780.0-0.43850.00.26420.0yy
88.702-0.20010.00.36150.0-0.46220.00.340.00.00.00.340.0-0.46220.00.36150.0-0.2001yy
101.7220.0-0.07790.00.21240.0-0.39750.00.53930.0-0.53930.00.39750.0-0.21240.00.07790.0yy
101.7270.03980.0-0.13640.00.3030.0-0.4810.00.55960.0-0.4810.00.3030.0-0.13640.00.0398yy

Table 2. The energy, wave functions and magnetic easy axis of 17 CEF singlets.

Note that the CEF splitting would induce magnetic anisotropy. To see this, we calculate the 3×33 \times3 matrix Mi,j≡⟨φ∣J^iJ^j∣φ⟩M_{i,j} \equiv\langle\varphi\vert\hat{J}_i\hat{J}_j\vert\varphi\rangle for i,j=x,y,zi,j = x,y,z and φ\varphi representing the CEF states. The magnetic easy axis of φ\varphi is determined as the eigenvector corresponding to the largest eigenvalue of Mi,jM_{i,j} [43]. As a result, the ground state of the Ho3+^{3+} ion exhibits Ising anisotropy, with its easy axis lying in the abab plane. Moreover, as shown in Table 2, we find that CEF excitations may be accompanied by spin-flip. For these spin-flip states, in addition to the CEF energy, the energy contribution from the Heisenberg exchange interaction is also included. Specifically, states with an easy axis along the yy or zz direction possess an additional magnetic exchange energy of J2S2J_2S^2 relative to the ground state. Using [13] J2=0.23J_2 = 0.23 meV and S=2S = 2, this energy contribution is calculated to be 0.92 meV. Furthermore, our results show that the highest CEF level, energetically 101.727 meV above the ground state, features an in-plane easy axis perpendicular to that of the ground state. This aligns closely with the measured energy barrier of 107 meV associated with the spin-flip process.

Orbitally resolved density of states for Ag 4d, Ge 4p, and Ho 4f

Figure 5 (PDF p. 13).

Extended STM data analysis

Fourier transform analysis

Figure S2. Fourier transform analysis. We show the Fourier transform analysis for the map data set in the area of Figure 3, which include the Fourier transform of topographic data (left), map data at the dip energy (middle) and map data at zero-energy (right). In all three cases, the clearest signal is the Bragg peaks marked by red circles, indicating the major spatial modulation of the electronic states comes from the lattice. The vertical stripes are from the instrumental noise. Add data are taken at 300mK.

Additional linecut spectrums

Figure S3. Additional linecut spectrums (b) taken on the position marked on the topographic image in (a). Add data are taken at 300mK.

References

  1. [1]C. Broholm, R. J. Cava, S. A. Kivelson, D. G. Nocera, M. R. Norman, and T. Senthil, Science 367, eaay0668 (2020).
  2. [2]C. Nisoli, R. Moessner, and P. Schiffer, Rev. Mod. Phys. 85, 1473 (2013).
  3. [3]S. H. Skjærvø, C. H. Marrows, R. L. Stamps, and L. J. Heyderman, Nat. Rev. Phys. 2, 13 (2020).
  4. [4]J. X. Yin, B. Lian, and M. Z. Hasan, Nature 612, 647 (2022).
  5. [5]Y. Wang, H. Wu, G. T. McCandless, J. Y. Chan, and M. N. Ali, Nat. Rev. Phys. 5, 635 (2023).
  6. [6]L. Pauling, J. Am. Chem. Soc. 57, 2680 (1935).
  7. [7]M. J. Harris, S. T. Bramwell, D. F. McMorrow, T. Zeiske, and K. W. Godfrey, Phys. Rev. Lett. 79, 2554 (1997).
  8. [8]A. P. Ramirez, A. Hayashi, R. J. Cava, R. Siddharthan, and B. S. Shastry, Nature 399, 333 (1999).
  9. [9]S. T. Bramwell and M. J. P. Gingras, Science 294, 1495 (2001).
  10. [10]C. Castelnovo, R. Moessner, and S. L. Sondhi, Nature 451, 42 (2008).
  11. [11]D. J. P. Morris et al., Science 326, 411 (2009).
  12. [12]T. Fennell, P. P. Deen, A. R. Wildes, K. Schmalzl, D. Prabhakaran, A. T. Boothroyd, R. J. Aldus, D. F. McMorrow, and S. T. Bramwell, Science 326, 415 (2009).
  13. [13]K. Zhao, H. Deng, H. Chen, K. A. Ross, V. Petříček, G. Günther, M. Russina, V. Hutanu, and P. Gegenwart, Science 367, 1218 (2020).
  14. [24][24]P. A. M. Dirac, Proceedings of the Royal Society of London. Series A, Containing Papers of a Mathematical and Physical Character 133, 60 (1931).
  15. [25][25]A. J. Heinrich, J. A. Gupta, C. P. Lutz, and D. M. Eigler, Science 306, 466 (2004).
  16. [26][26]C. F. Hirjibeh edin, C.-Y. Lin, A. F. Otte, M. Ternes, C. P. Lutz, B. A. Jones, and A. J. Heinrich, Science 317, 1199 (2007).
  17. [27][27]X. Chen et al., Phys. Rev. Lett. 101, 197208 (2008).
  18. [28][28]T. Miyamachi et al., Nature 503, 242 (2013).
  19. [29][29]M. Ternes, New Journal of Physics 17, 063016 (2015).
  20. [30][30]T. Balashov, A. F. Takács, W. Wulfhekel, and J. Kirschner, Phys. Rev. Lett. 97, 187201 (2006).
  21. [31][31]Somesh Chandra Ganguli, Markus Aapro, Shawulienu Kezilebieke, Mohammad Amini, Jose L. Lado, and Peter Liljeroth, Nano Letters 23(8), 3412-3417 (2023).
  22. [32][32]K. Zhao et al. Nat. Phys. 20, 442–449 (2024).
  23. [33][33]W. H. A. Thijssen, D. Djukic, A. F. Otte, R. H. Bremmer, and J. M. van Ruitenbeek, Phys. Rev. Lett. 97, 226806 (2006).
  24. [34]See Supplemental Material [url] for the calculation of the surface crystal electric field, which includes Refs. [39-43].
  25. [35]R. Wiesendanger Rev. Mod. Phys. 81, 1495 (2009).
  26. [36][36]J.-X. Yin et al., Nat. Commun. 11, 4415 (2020).
  27. [37][37]Y. Xing et al., Nat. Commun. 11, 5613 (2020).
  28. [38][38]J.-X. Yin et al., Phys. Rev. Lett. 129, 166401 (2022).
  29. [39][39]G. Kresse and J. Furthmuller, Phys Rev B Condens Matter 54, 11169 (1996).
  30. [40][40]J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett. 78, 1396 (1997).
  31. [41][41]D. J. Newman and B. Ng, Crystal Field Handbook (Cambridge University Press, Cambridge, 2000).DOI
  32. [42][42]A. Scheie, J. Appl. Crystallogr. 54, 356 (2021).
  33. [43][43]S. Watanabe, Proceedings of the National Academy of Sciences 118, e2112202118 (2021).

Paper details

vlnc:W_py4pjbs6
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents