2025 Volume 38 Issue 4
Article Contents

Jiaan Cao, Lyuzhou Ye, Rui-Xue Xu, Xiao Zheng. Influence of Radio-Frequency Voltage on Electron Spin Resonance Spectroscopy in Scanning Tunneling Microscopy†[J]. Chinese Journal of Chemical Physics, 2025, 38(4): 375-381. doi: 10.1063/1674-0068/cjcp2409127
Citation: Jiaan Cao, Lyuzhou Ye, Rui-Xue Xu, Xiao Zheng. Influence of Radio-Frequency Voltage on Electron Spin Resonance Spectroscopy in Scanning Tunneling Microscopy†[J]. Chinese Journal of Chemical Physics, 2025, 38(4): 375-381. doi: 10.1063/1674-0068/cjcp2409127

Influence of Radio-Frequency Voltage on Electron Spin Resonance Spectroscopy in Scanning Tunneling Microscopy†

通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Figures(4)

Article Metrics

Article views(242) PDF downloads(0) Cited by(0)

Access History

Influence of Radio-Frequency Voltage on Electron Spin Resonance Spectroscopy in Scanning Tunneling Microscopy†

Abstract: Over the last decade, the integration of scanning tunneling microscopy (STM) and electron spin resonance (ESR) spectroscopy has emerged as a powerful tool for measuring spin states of surface-adsorbed molecules. The radio-frequency voltage is a key physical quantity that influences STM-ESR spectra. However, the specific effect of radio-frequency voltage on the real-time electric current associated with STM-ESR signal remains unclear. In this work, we employ the hierarchical equations of motion method to simulate the STM-ESR spectra of a single spin-1/2 surface-adsorbed molecule and track the temporal evolution of the electric current, thereby elucidating how the radio-frequency voltage influences the features of STM-ESR spectra, the real-time electric current, and the characteristic frequencies conveyed by the electric current. These theoretical insights facilitate a deeper comprehension of experimental phenomena.

    • In the past few decades, single magnetic molecules adsorbed on surfaces have received widespread attention due to their potential applications in spintronics [1–3], data storage [4–6], quantum information [1, 7, 8], and quantum computation [1, 5, 9]. The crucial aspect for realizing these applications lies in the precise measurement and control of molecular spin states. The synergistic integration of scanning tunneling microscopy (STM) and electron spin resonance (ESR) spectroscopy, known as STM-ESR, has recently become a powerful and advanced tool for probing and analyzing spin states in magnetic molecules [1–5, 7–27]. It combines the atomic-scale spatial resolution of STM (sub-angstrom) and the ultra-high energy resolution of ESR (nano-electronvolt), enabling unprecedented insights into discerning spin excitations and spin-spin interactions within surface-adsorbed magnetic molecules [1–3, 9–14, 20–22].

      FIG. 1 illustrates a schematic of the setup typically used in the STM-ESR experiments, where a single localized spin-1/2 molecule (e.g. TiH) is deposited on an insulating substrate (e.g. double-layer MgO on Ag). To enable high-resolution spin sensing, the STM tip is spin-polarized by attaching magnetic atoms (e.g. Fe) to its apex [1–3, 9–12, 28]. An external static magnetic field ($ \sim1\,{\rm{T}} $), denoted as $ {{\boldsymbol{B}}}_{\rm{ext}} $, is applied to induce Zeeman splitting of molecular spin energy levels. The STM-ESR signal is obtained as the direct current (dc) component of the real-time electric current $ I(t) $ driven by a time-dependent bias voltage $V(t) = V_{\rm{rf}}{\rm{sin}}(\omega_{\rm{rf}}t)+V_{\rm{dc}}$ [3, 10, 11]. Here, $ V_{\rm{dc}} $ represents a constant dc voltage, while $ V_{\rm{rf}} $ and $ \omega_{\rm{rf}} $ denote the amplitude and frequency of a continuous-wave radio-frequency (rf) voltage, respectively.

      Numerous experimental studies have reported that $ V_{\rm{rf}} $ is one of key physical quantities that affect the features of STM-ESR spectra and Rabi frequency [4, 7, 10, 16, 17, 29]. For instance, Willke et al. have found that the peak height of the STM-ESR spectra of a single Fe atom is positively correlated with $ V_{\rm{rf}} $ until it reaches saturation [17]. Seifert et al. have investigated the dependence of the amplitude and full width at half maximum (FWHM) of STM-ESR spectra on $ V_{\rm{rf}} $ for a single Fe atom and a TiH molecule [29]. They observed that the FWHM grows almost linearly with $ V_{\rm{rf}} $. This observation has also been confirmed by Bae and coworkers through similar experiments on a strongly coupled TiH dimer [16]. Moreover, Willke et al. and Yang et al. have demonstrated the Rabi oscillations of individual TiH molecules and iron phthalocyanine molecules at different $ V_{\rm{rf}} $, respectively [7, 10]. Their results consistently show that the Rabi frequency is linearly proportional to $ V_{\rm{rf}} $.

      Despite a variety of experimental evidence demonstrating the significant impact of $ V_{\rm{rf}} $ on STM-ESR spectra and Rabi oscillations, further exploration in this field still faces several challenges: First, the STM-ESR spectra result from averaging the long-time electric current. However, it is difficult to directly probe the real-time electric current associated with STM-ESR spectra in experiments due to the time resolution limited by the electronic response time [30], thereby the specific effect of $ V_{\rm{rf}} $ on the real-time electric current remaining unknown. Second, a few theoretical studies have explored the variation of STM-ESR spectra with respect to $ V_{\rm{rf}} $ [25, 31]. However, in these studies, the FWHM of the calculated STM-ESR spectra is independent of $ V_{\rm{rf}} $, which does not agree with the experimental observations. This may be attributed to the limitations of the theoretical methods used in accuracy. Thus, further theoretical investigations are necessary.

      The numerically exact hierarchical equations of motion (HEOM) method implemented in the general-purpose program HEOM-QUICK2 has emerged as a promising and useful tool to tackle the aforementioned challenges [28, 32–37]. It is capable of effectively handling arbitrary time-dependent external fields and accessing the evolution information of system observables in real time. In particular, it has been successfully applied to simulate the STM-ESR spectra of a single TiH molecule and elucidate the fundamental origin of the STM-ESR signals [22].

      In this work, we employ the HEOM method to simulate the STM-ESR spectra of a single spin-1/2 surface-adsorbed molecule at different $ V_{\rm{rf}} $ and track the temporal evolution of electric current, which aims at elucidating the influence of $ V_{\rm{rf}} $ on the features of the STM-ESR spectra, the real-time electric current, and the characteristic frequencies conveyed by the electric current.

    II.   MODEL AND METHODOLOGY
    • Several first-principles-based calculations [38–40] have demonstrated that the single-orbital Anderson impurity model (AIM) [41, 42] can be utilized to describe the STM tunnel junction illustrated in FIG. 1. The total Hamiltonian is comprised of three parts (we set $ \hbar = k_{\rm{B}} = e = 1 $ hereafter), i.e., $ H_{\rm{T}} = H_{\rm{imp}}+ H_{\rm{env}}+ H_{\rm{coup}} $. The individual spin-1/2 surface-adsorbed molecule is represented by a single impurity orbital, with its Hamiltonian described by

      Here, ${\hat n}_{\sigma} = {\sum}_{\sigma}{\hat d}_{\sigma}^\dagger{\hat d}_{\sigma}$ is the spin-$ \sigma $ ($\sigma=\uparrow,\downarrow$) electron number operator, and $ \epsilon $ and $ U $ are the orbital energy and the electron-electron Coulomb repulsion energy, respectively. $ \hat{\boldsymbol{S}} $ is the local spin operator, $ g $ is the gyromagnetic ratio, $ \mu_{\mathrm{B}} $ is the Bohr magneton, and $ {{\boldsymbol{B}}}_{\rm{ext}} $ is the external magnetic field that includes the field generated by the permanent spin moment of the magnetic atoms at the tip apex. Notably, the $ g\mu_{{\mathrm{B}}}\vert{{\boldsymbol{B}}}_{\rm{ext}}\vert $ represents the Zeeman energy, which corresponds to the Larmor frequency $ \omega_{\rm{L}} $.

      In particular, the environment including the tip (t) and the substrate (s) is described by two noninteracting electron reservoirs, i.e., $ H_{\rm{env}}=\sum_{\alpha=\mathrm{t,s}}\sum_{k\sigma}\epsilon_{\alpha k\sigma} \hat{c}_{\alpha k\sigma}^{\dagger}\hat{c}_{\alpha k\sigma} $, where $ {\hat c}^\dagger_{\alpha k \sigma} $ ($ {\hat c}_{\alpha k \sigma} $) is the creation (annihilation) operator for a spin-$ \sigma $ electron on the $ k $th band in $ \alpha $-reservoir (α=t, s). The impurity-environment coupling takes a linear form of $H_{\rm{coup}} ={\sum}_{\alpha k \sigma}a_{\alpha k \sigma}{\hat d}_{\sigma}^\dagger{\hat c}_{\alpha k \sigma}+ H.c.$, with $\{a_{\alpha k \sigma}\}$ being the coupling strengths between the impurity orbital and the $ k $th band of $ \alpha $-reservoir. Since the noninteracting electron reservoirs follow Gaussian statistics, their influence on the impurity is completely manifested by the hybridization functions, $ \Gamma_{\alpha \sigma}(\omega) = \Gamma_{\alpha\sigma}D_\alpha (\omega) $. Here, $ \Gamma_{\alpha\sigma} $ and $D_\alpha (\omega) $ represent the hybridization strength and dimensionless density of states of $ \alpha $-reservoir, respectively. For numerical convenience, $ D_\alpha (\omega) $ adopts a simple Lorentzian form, i.e., $ D_{\alpha}(\omega)=W_{\alpha}{^2}/[(\omega-\mu_{\alpha})^2+W_{\alpha}{^2}] $, where $ \mu_\alpha\left(W_{\alpha}\right) $ is the band center (width) of $ \alpha $-reservoir, and $ \mu_\alpha $ is also the chemical potential of $ \alpha $-reservoir. The tip spin polarization is oriented along the $ {\hat{\boldsymbol{z}}} $ direction, and its extent is quantified by $ \chi $, defined as $ \chi\equiv(\Gamma_{\mathrm{t}\uparrow}-\Gamma_{\mathrm{t}\downarrow})/\Gamma\mathrm{_t} $, with $ \Gamma\mathrm{_t}=\Gamma_{\mathrm{t}\uparrow}+\Gamma_{\mathrm{t}\downarrow} $.

      The parameters adopted in the AIM are determined based on experimental data [3, 11] and first-principles calculations [38–40]. Specifically, the value of the orbital energy $ \epsilon $ is obtained from the experimentally measured differential conductance spectrum reported in Ref.[3], which is consistent with the density functional theory calculation results of Ref.[38]. The Coulomb repulsion energy $ U $ is determined from the first-principles calculations presented in Ref.[39]. Other energy parameters, such as temperature $ T $, voltage amplitudes $ V_{\rm{dc}} $ and $ V_{\rm{rf}} $, Zeeman energy $g\mu_{\rm B}\vert{{\boldsymbol{B}}}_{\rm{ext}}\vert$, and hybridization strength $ \Gamma_{\alpha\sigma} $, are extracted from experimental data [3, 11]. This careful selection of parameters ensures that our AIM reliably represents the realistic experimental setup, offering a solid foundation for precise simulations of the STM-ESR spectra.

      All simulations in this work are performed using the AIM. In this model, the applied bias voltage $ V(t) $ drives both charge and spin dynamics within the STM tunnel junction. It is noted that the energy scale associated with spin dynamics $ (\sim 0.1\,{\rm{meV}}) $ is significantly lower than that of charge dynamics $ (\sim 1\,{\rm{eV}}) $. This substantial disparity poses a major obstacle for numerical simulations, that is, accurate time propagation of the HEOM demands an extremely small time step due to the energy-time uncertainty relation, leading to computationally intensive simulations, especially when extracting the dc component of the electric current. To address this difficulty, we scale down the energetic parameters associated with charge-state excitations, such as $ \epsilon $, $ V_{\rm{dc}} $ and $ V_{\rm{rf}} $, by the same factor, while maintaining a clear distinction between charge and spin energy scales. Other energetic parameters, such as $ T $, $g\mu_{\rm B}\vert{{\boldsymbol{B}}}_{\rm{ext}}\vert$ and $ \Gamma_{\alpha\sigma} $, are carefully selected based on experimental values [3, 11]. The values of all these parameters adopted in numerical simulations are detailed in the captions of figures. The real-time electric current is obtained from the first-tier auxiliary density operators calculated using the HEOM method [35, 37].

    III.   RESULTS AND DISCUSSION
    • FIG. 2(a) depicts the simulated STM-ESR spectra of a single spin-1/2 surface-adsorbed molecule at different $ V_{\rm{rf}} $, which are obtained by

      During the measurement, the frequency $ \omega_{\rm{rf}} $ of the rf voltage is continuously swept over a range of values. The measured STM-ESR spectra exhibit resonance at the frequency $ \omega_{\rm{L}} $, which corresponds to the Zeeman energy between the two spin eigenstates ($ \vert0\rangle $ and $ \vert1\rangle $) in the singly-occupied subspace; see the inset. Here, $ \vert0\rangle $ ($ \vert1\rangle $) indicates that the local spin aligns parallel (antiparallel) to $ {{\boldsymbol{B}}}_{\rm{ext}} $. The spectral resonance features can be diverse, resulting in both symmetric and asymmetric resonance peaks. As $ V_{\rm{rf}} $ increases, the resonance peaks gradually transform from a symmetric Lorentzian lineshape to an asymmetric lineshape, closely resembling the experimental results reported in Refs.[2, 10]. The underlying mechanisms that govern the essential features of the STM-ESR spectra have been clarified in our previous work; see Ref.[22] for more details.

      Moreover, all resonance peaks in FIG. 2(a) are scaled up to the same height and vertically offset for better visibility. It is found that the peak height shows a nonlinear positive correlation with $ V_{\rm{rf}} $ [17, 22]. The FWHM of the simulated spectra is almost linearly proportional to $ V_{\rm{rf}} $, as illustrated in FIG. 2(b). These observations are in excellent agreement with experimental findings [2, 4, 17, 29], greatly validating the effectiveness and reliability of the AIM and the HEOM method.

      Importantly, we also track the real-time evolution trajectory of electric current $ I(t) $ associated with the STM-ESR signals, starting from when the system is in a steady state with the applied $ V_{\rm{dc}} $, as shown in FIG. 2(c). Here, the simulation results for $ V_{\rm{rf}} = 0.1, 0.5,1.5\,{\rm{mV}} $ are selected as representatives, and the rf frequency is precisely tuned to the resonance at $ \omega_{\rm{rf}}/2\pi = 24.12\,{\rm{GHz}} $, which is determined from the STM-ESR spectra in FIG. 2(a). Since the temporal evolution of $ I(t) $ contains a high-frequency oscillation illustrated in FIG. 2(d), its global view in FIG. 2(c) appears as a colored patch. It can be seen that the oscillation amplitude of $ I(t) $ progressively enhances with increasing $ V_{\rm{rf}} $, and its envelope displays a low-frequency oscillatory behavior. FIG. 2(d) correspondingly presents a magnified view of $ I(t) $ within a short time interval. These oscillations observed at different $ V_{\rm{rf}} $ have nearly identical high frequencies. Therefore, the temporal evolution of $ I(t) $ conveys two characteristic frequencies with significant numerical disparities.

      To clarify physical meaning of these two frequencies, we conducted the following theoretical analysis based on Ref.[22]. The charge and spin dynamics show a distinct separation in energy scales [1, 21, 22], and the phenomena of interest in this paper are primarily governed by spin dynamics, thus allowing for studies within the single occupation subspace. In this subspace, the AIM is converted to a spin-polarized Kondo model through the Schrieffer-Wolff transformation [43, 44], with the total Hamiltonian becoming the form of $\tilde H_{\rm{T}} = H_{\rm{spin}}+ H_{\rm{env}}+H_{\rm{int}}$. Here, $H_{\rm{spin}} = g\mu_{\rm B}{{\boldsymbol{B}}}_{\rm{ext}}\cdot{\hat{\boldsymbol{S}}}$ denotes the Hamiltonian of the local spin in the large-$ U $ limit. The quantum master equation describing the local spin dynamics can be written as [22]

      Here, $ \rho(t) $ is the system-reduced density matrix, and its relationship with the real-time electric current $ I(t) $ has been derived in our previous work [22], indicating the temporal evolution of electric current is fully determined by $ \rho(t) $. $ R(t) $ is the dissipative term that describes the impact of reservoir fluctuations on the spin dynamics, which accounts for the spin transfer torque mechanism described in Refs.[26, 31]. Due to the weak spin-reservoir coupling, $ R(t) $ can be neglected, and Eq.(3) formally recovers the dynamical theory of conventional ESR [45]. The $ H_{\rm{spin}} $ and $ {B}_{\rm{tip}}^{\rm{eff}}(t) $ determine the Larmor frequency and Rabi frequency, respectively. Thus, the characteristic frequencies conveyed by the real-time electric current correspond to these two frequencies. Specifically, the slower one manifested by the envelope of $ I(t) $ corresponds to the Rabi frequency (see FIG. 2(c)), while the faster oscillation frequency manifested by the magnified view of $ I(t) $ matches the Larmor frequency.

      The $ {B}_{\rm{tip}}^{\rm{eff}}(t) $ in Eq.(3) denotes an effective alternating magnetic field along $ {\hat{\boldsymbol{z}}} $. It manifests the mean-field effect originating exclusively from the spin-polarized tip, and has the form of $ {B}_{\rm{tip}}^{\rm{eff}}(t)\simeq B_0+B_1{\rm{sin}}(\omega_{\rm{rf}}t) $, where $ B_0 $ and $ B_1 $ are expressed as [22]

      Here, $ f_\beta (\omega) $ is the Fermi function at the inverse temperature $ \beta = 1/T $, Dt represent the dimensionless density of states of tip. The $ B_0 $ leads to a renormalized Zeeman splitting of spin energy levels, while $ B_1 $ serves as the driving source for the dynamics of local spin state and determines the Rabi frequency [22]. Notably, $ B_0 $ is not equivalent to the magnetic field generated by the permanent spin moment of the magnetic atoms at the tip apex in $\boldsymbol B_{\rm{ext}}$, which arises from the mean-field effect of the spin-polarized tip on the surface-adsorbed molecule and is essentially an effective magnetic field. Some studies have also mentioned the term “effective magnetic field” and attributed it to the piezoelectric effect [12, 23, 26]. However, $ {B}_{\rm{tip}}^{\rm{eff}}(t) $ discussed here differs from what they described, as it is generated by the oscillation of the electrochemical potential of the spin-polarized tip [22].

      To further investigate the influence of $ V_{\rm{rf}} $ on the Larmor frequency and Rabi frequency, we perform a half-Fourier transform on the real-time electric current using the following formula,

      The transformed results are displayed in FIG. 3. The frequency-domain representation ${\cal{I}}(\omega)$ of the real-time electric current exhibits only two characteristic peaks that differ in frequency by approximately 2 to 3 orders of magnitude. FIG. 3(a) depicts the high-frequency peaks of $ {\cal{I}}(\omega) $ at different $ V_{\rm{rf}} $, with their positions corresponding to the oscillation frequency in FIG. 2(d), i.e., the Larmor frequency $ \omega_{\rm{L}} $. Since $ \omega_{\rm{L}} $ is primarily determined by $ \vert{{\boldsymbol{B}}}_{\rm{ext}}\vert $ and is independent of $ V_{\rm{rf}} $, the positions of the high-frequency peaks are almost the same at different $V_{\rm{rf}}$, with variation only in the peak amplitude. The other characteristic peaks with lower frequencies in $ {\cal{I}}(\omega) $ are illustrated in FIG. 3(b), and their positions match the oscillation frequency in FIG. 2(c), i.e., the Rabi frequency $ \Omega $. In comparison to the high-frequency peaks, the amplitude of the low-frequency peaks is also positively correlated with $ V_{\rm{rf}} $, but their positions demonstrate a pronounced dependence on $ V_{\rm{rf}} $. This implies that $ V_{\rm{rf}} $ is a key factor in controlling the Rabi frequency of a single spin-1/2 surface-adsorbed molecule.

      We also explore how $ V_{\rm{rf}} $ influences the Rabi frequency through both numerical simulations and analytical calculations, as shown in FIG. 4. The data points, represented by circles, are determined from those low-frequency peaks shown in FIG. 3(b), and the red dashed line is a linear fit to these points. It is evident that the Rabi frequency $ \Omega $ increases linearly with $ V_{\rm{rf}} $. Note that our simulations may fail when $ V_{\rm{rf}} $ is very small, as in this case the charge and spin energy scales are not well separated. Hence, we do not consider the cases with very small $ V_{\rm{rf}} $ here. Furthermore, based on the analytic theory developed in Ref.[22], we also calculate the Rabi frequency $ \Omega^\prime $ at these $ V_{\rm{rf}} $ using Eq.(5), with its expression given by $\Omega^\prime = g\mu_{\rm B} B_1/h$. The corresponding calculation results are displayed by the rectangles in FIG. 4, and the blue dashed line represents their linear fit. Although the simulated Rabi frequency ($ \Omega $) is not exactly equal to the calculated value ($ \Omega^\prime $) due to approximations made in both ways, their difference is not significant and both exhibit a linear dependence on $ V_{\rm{rf}} $. This is highly consistent with the experimental findings [7, 9, 10]. Therefore, our work provides a mutual validation of the experimental, calculated, and simulated results, collectively elucidating how $ V_{\rm{rf}} $ affects the Rabi frequency.

    IV.   CONCLUSION
    • In summary, we have utilized the HEOM method to simulate the STM-ESR spectra of a single spin-1/2 surface-adsorbed molecule. The FWHM of the resulting spectral peaks increases almost linearly with $ V_{\rm{rf}} $, which aligns well with experimental observations. This greatly validates the effectiveness and reliability of the AIM and the HEOM method. More importantly, our simulations have traced the temporal evolution of the electric current associated with the STM-ESR spectra, and revealed two characteristic frequencies conveyed by the real-time electric current, i.e., the Larmor frequency and Rabi frequency. The Larmor frequency is independent of $ V_{\rm{rf}} $ and is primarily determined by $ \vert{{\boldsymbol{B}}}_{\rm{ext}}\vert $, while the Rabi frequency is linearly proportional to $ V_{\rm{rf}} $. This relationship is validated through both numerical simulations and analytical calculations. The theoretical insights presented in this work contribute to a better understanding of experimental phenomena and provide an important basis for the validation of experimental protocols.

    Figure (4)  Reference (45)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return