Skip to main content
NCBI home page
Scientific Reports logoLink to Scientific Reports
. 2015 Mar 10;5:8943. doi: 10.1038/srep08943

Controllable magnetic correlation between two impurities by spin-orbit coupling in graphene

F M Hu 1,a, Liangzhi Kou 1,b, Thomas Frauenheim 1
PMCID: PMC4354095  PMID: 25754911

Abstract

Two magnetic impurities on the edge of a zigzag graphene nanoribbon strongly interact with each other via indirect coupling, which can be mediated by conducting carriers. By means of Quantum Monte Carlo (QMC) simulations, we find that the spin-orbit coupling λ and the chemical potential μ in system can be used to drive the transition of local-spin exchange from ferromagnetism to anti-ferromagnetism. Since the tunable ranges for λ and μ in graphene are experimentally reachable, we thus open the possibilities for its device application. The symmetry in spatial distribution is broken by the vertical and the transversal spin-spin correlations due to the effect of spin-orbit coupling, leading to the spatial anisotropy of spin exchange, which distinguish our findings from the case in normal Fermi liquid.

Due to the unusual electronic quasi-particles and the flexible carrier concentration1, graphene with local spins is regarded as a novel type of two-dimensional magnetic semiconductor as well as a promising candidate for spintronics devices2. Doping with magnetic impurities is regarded as an effective route to induce local spins in graphene3, because the pristine graphene is a nonmagnetic semimetal without band gap. In previous works concerning single-impurity doped graphene, it was reported that applied gate voltage (to change the chemical potential) can control the electron filling on impurity orbit4,5 and its Kondo screening from conducting electrons6,7, thus leading to an on-off switching for a single spin in graphene. But in realistic spin-based devices, we need to consider multi-spin case and spin-flip process owing to the experimental challenge of doping concentration control, which will induce spin current transport in graphene systems. Therefore, for the device application, it is crucial to achieve efficient controllability for spin correlations among embedded impurities, which is measured by the Ruderman-Kittel-Kasuya-Yosida (RKKY) interaction and mediated by conducting electrons. In previous studies on RKKY interaction, the electron doping dominates the large-distance oscillation with a wave vector 2kF as well as decay behavior R−γ 8.

Recently, there is a surge of interest in effect of spin-orbit coupling (SOC) on electronic properties, which opens a class of novel time-reversal invariant topological insulators (TI)9. On the edge (two-dimension case) or surface (three-dimension case) of these TI materials, the carries have spin-momentum locking, so called “helical” states: electrons with opposite momenta have orthogonal spin orientations, so that the backscattering k ↔ −k is forbidden without a spin flip. Due to this scenario, there is spin-current vortex around the single local spin on the helical edge10. Furthermore, on the surface of three-dimensional realistic TI, Kondo impurities will open up a local gap11 and the exchanges can be described by the RKKY interaction among the impurities12. More interestingly, the combination of anisotropic spin exchange and SOC on the edge of two-dimensional realistic TI results in unnormal temperature-dependent conductance on the edge13,14. These investigations indicate that the SOC can greatly modify the spin configuration for single impurity in realistic TI, such as class of Bi2Se3 (Bi2Te3, Sb2Te3)15 and strained HgTe16. While in graphene, which is one of the earliest proposals for TI materials, multi-spin problem has not been comprehensively studied, despite the facts that the SOC in graphene can be enhanced in a wide range, while both the Kondo screening around single spin and RKKY interaction between two spins can be greatly affected by the SOC.

In this paper, we consider two Anderson impurities on the edge of a zigzag graphene ribbon, and investigated the effect of Kane-Mele SOC and chemical potential on the local spin exchange, which can be achieved by heavy metal doping or electric field. Through Quantum Monte Carlo simulation at finite temperature, we carry out a comprehensive study on the spin and charge coupling strengthes of this impurity pair. For the spin-spin correlation, which measures the RKKY interaction, we found that, the in-plane components of correlations favor ferromagnetism but the out-plane correlation can be tuned from ferromagnetism to antiferromagnetism by the SOC, while such transition is much dependent on the on-site Coulomb interaction of impurity orbital. From recent first-principal and experimental results of enhanced SOC in graphene, we conclude that spin-spin exchange among impurities in graphene is experimentally tunable by SOC. In contrast, the charge fluctuation between two impurities is always suppressed. Meanwhile, in-plane components of correlation will decay with an oscillation behavior when the distance between two impurities is increased, but the out-plane spin and charge correlations will decay smoothly, which is totally different from the Fermi liquid behavior, and leading to the spatial anisotropy of spin exchange.

To simulate the model, we construct a model with width of 20 Å zigzag-edge graphene ribbon and along the horizontal direction we adopt periodical condition. As indicated in Kane-Mele investigation17, graphene will be driven into topological phase when the intrinsic SOC in graphene is enhanced. In the case, the ribbon would possess the helical edge states. In Fig. 1, we show the scheme for two impurities (yellow balls) located on the zigzag helical edge, where red and blue arrows denote the conducting electrons with up and down spins, respectively, and they propagate along oppositive directions (spin-momentum locking).

Figure 1. The scheme for the two magnetic impurities (yellow balls) at the edge of the zigzag nanoribbon.

Figure 1

Open in a new tab

The arrows on the top and bottom demotes the moving directions for two species spins.

The QMC simulation is based on two-Anderson-impurity model, the total Hamiltonian reads as

graphic file with name srep08943-m1.jpg

where HK-M is the Kane-Mele model in a zigzag edge graphene ribbon17. It is composed with two parts: HK-M = Ht + Hso, with Ht being the usual nearest-neighbor hopping of tight-binding model in graphene

graphic file with name srep08943-m2.jpg

and

graphic file with name srep08943-m3.jpg

Hso is the SOC term and σz is the z Pauli matrix. Hso thus has opposite signs for opposite electron spins. The parameters νij = −νji = ±1 depend on the orientation of the two nearest neighbor bonds as the electron hops from site i to j: νij = +1 if the electron makes a left turn to the second bond. It is negative if it makes a right turn. H1 is the part for two impurities

graphic file with name srep08943-m4.jpg

and H2 describes the hybridization between impurities and conducting electrons

graphic file with name srep08943-m5.jpg

Here m is the index of two impurities, in H1 and H2, εm is the energy of the impurity orbital and Um is Coulomb repulsion inhibiting the simultaneous occupancy of the orbital by two electrons and Vm is the hybridization strength. In this work, we consider the two identical impurities, so the above parameters in Hamiltonian are m-independent.

We use the single-impurity QMC base on discrete-time algorithm18 to compute the local thermodynamic properties of the impurity. The length of imaginary-time is set as Δτ = 0.1 to suppress the systemic error. The QMC naturally returns the imaginary-time Green's function Inline graphic of the impurity. With this Green's function, we can compute the spin-spin correlations for two impurities along α = x, y, z directions versus distance as

graphic file with name srep08943-m6.jpg

where Inline graphic, Inline graphic and Inline graphic; as well as the charge-charge correlation as the function of distance

graphic file with name srep08943-m7.jpg

and Inline graphic. The spin-spin correlations Sα (R) directly reflect the RKKY interaction which is mediated by conducting electrons on the helical edge, while charge-charge correlation C(R) indicates the indirect Coulomb repulsion.

In Fig. 2, we show the spin-spin and charge-charge correlations as function of SOC, Coulomb interaction, and chemical potential. Here, we fixed the hole-particle symmetry μ = 0 and Inline graphic, so that Inline graphic and the local moments at two impurity sites are well developed. One can see from Fig. 2(a), the out-plane correlations of Sz highly depend on SOC, which can be reduced in a linearly way and have a wide variation range. When SOC is increased, the value of vertical correlation Sz can be changed from positive to negative, it means that a transition from a ferromagnetic (FM) to anti-ferromagnetic (AFM) correlation is achieved when the SOC reaches a critical value. It is interesting to note that, the critical SOC strength for such a transition increases as Coulomb interaction is increased, from the QMC simulations. For example, the critical SOC value is about 100 meV when U is 1 eV, but it is increased to 300 meV when U is is 5 eV. The underlying mechanism can be attributed to the competition of spin and charge fluctuations between two impurities. When they are close to each other, the changing in carriers SOC can directly result in the spin-flip process which reflects in Sz, but the increasing Coulomb interaction U can block spin flip, so it is seen that with a larger U in Fig. 2(a), the conducting electrons need larger kinetic energies (SOC) to drive a FM-AFM transition. But when two impurities are far away from with each other, we can not see this transition in Sz but in Sx(y), which is totally determined by the Luttinger liquid physics on the edge19. The explanation can be also demonstrated in the latter plots [Fig. 3(a)].

Figure 2.

Figure 2

Open in a new tab

(a) The spin-spin correlation along z direction Sz versus λ, (b) The spin-spin correlation along x or y direction Sx(y) versus λ (c) The charge-charge correlation versus λ and in (a)–(c) the inverse of temperature β = 32 t−1, V = 0.65 t, μ = 0 and Inline graphic. (d) Sz and Sx(y) versus chemical potential μ with β = 32 t−1, U = 1.2 t, V = 0.65 t and Inline graphic.

Figure 3.

Figure 3

Open in a new tab

(a) The vertical and transversal spin-spin correlation as function of distance between two impurities (b) The charge-charge correlation as function of distance. (c) The ratio Sx(y)(R)/C(R) with fixed chemical potential μ = −0.2 t and varied SOC λ = 0.1 t and 0.15 t. (d) The ratio Sx(y)(R)/C(R) with fixed SOC λ = 0.1 t and varied chemical potential μ = 0.1 t and 0.2 t. The x axis is in the unit of lattice constant of graphene.

The SOC driven magnetism transition will be important for the spintronics application. It is worthy to point out the required critical SOC is in a reasonable range based on the fact that recent extensive strategies were proposed and developed to enhance SOC in graphene. In Table 1, we sumarize the recent achievements from both density functional theory and experiments. In the case of interface/intercalation, the recent tunable range for SOC is 50–225 meV through placing graphene on the substrates of Au(111), Ag(111), and Ni(111) and intercalation with SiC (0001)20,21,22,23,24. Alternatively, doping graphene with heavy or magnetic adatoms (Ni, Au, Ti and Os) is also able to induced the enhanced SOC in graphene of 20–500 meV25,26,27. Meanwhile, when sandwiching graphene between strong spin orbit interactions or Rashba materials by the proximity effects, the intrinsic SOC is significantly enhanced and graphene is driven into topological phase with nontrivial gap of 120 meV28,29. Due to these efficient strategies, it is possible to tune the SOC in graphene over a wide range, the FM-AFM transition for spin correlation between two magnetic impurities via SOC controlling is thus feasible.

Table 1. Enhanced Gap ΔESOC in Graphene.

Method ΔSOC (meV)
Interface/intercalation 50–225 [20,21,22,23,24]
Adatom doping 20–500 [25,26,27]
Proximity effect 50–120 [28, 29]
Open in a new tab

In contrast to the spin-spin correction transition of Sz, the transversal correlation Sx(y) is increased linearly by the SOC interaction but not change its sign. In Fig. 2(b), we see that at SOC λ = 0, Sz = Sx(y), but this symmetry is broken when SOC is switched on. The transversal correlation Sx(y) is increased when either of SOC and Coulomb interaction is enhanced. More interestingly, the enhance SOC not only can change the spin correction between two impurities, but also affect the charge interaction between them. As indicated in Fig. 2(c), the charge-charge correlation as a function of SOC with different values of Coulomb interaction is calculated. We can see that the absolute value of C is increased when SOC is increased, but decreased as the Coulomb interaction is enhanced. When U is at 1.2 t, the charge fluctuation is totally suppressed. To have a comprehensive physics picture, we also vary chemical potential μ to break the hole-particle symmetry and study their effects on Sz and Sx(y), In Fig. 2(d). The vertical and transversal correlations are totally degenerated (i.e. Sz is equal to Sx(y)) at SOC strength λ = 0. But when λ ≠ 0, the SOC effect on both spin correlation is different, leading to non-degenerated Sz and Sx(y). We then turn to the effect of chemical potential on Sz and Sx(y), it could be found that both Sz and Sx(y) are positive near the zero point of μ, suggesting that two impurities correlate ferromagnetically. After μ is shifted from 0, the two impurities have anti-ferromagnetic correlation as indicated by the negative values of Sz and Sx(y) (μ is in the range of −0.2 t to −0.8 t). But when μ is further reduced, both Sz and Sx(y) approach to zero, indicating the weak coupling between two impurities.

The distance between two impurities is expected to influence the couplings of spin-spin and charge-charge due to the fact that the RKKY interation is sensitive to the distance. We preformed the QMC calculations to address the problem when varying the distance. In Fig. 3, we show the spin-spin and charge-charge correlations as functions of distance between two impurities. Here we consider the case that both of two impurities are located on the edge (the same sub-lattice). In Fig. 3(a), we see that the vertical spin-spin correlation Sz(R) and transversal one Sx(y)(R) have different spatial distribution. Sx(y)(R) has both the behavior of decaying and oscillating, but Sz(R) is always decaying with the increased distance. Similar to Sz(R), the charge-charge correlation is also decaying with distance (Fig. 3(b)). These correlation functions have the following behavior:

graphic file with name srep08943-m8.jpg
graphic file with name srep08943-m9.jpg

and

graphic file with name srep08943-m10.jpg

here f1(R), f2(R) and f3(R) are decaying function of distance has the formula as 1/Rγ.

To more clearly display these correlations, we calculate the ratios Sx(y)(R)/C(R) and Sz(R)/C(R). We find that Sz(R)/C(R) is a trivial constant, but Sx(y)(R)/C(R) is an oscillating function of distance, as shown in Fig. 3 (c) and (d). The ratio exhibits prefect behavior of sine or cosine functions with wave vector, which depends on the chemical potential and SOC.

Finally, we discuss the model Hamiltonian of Eq. (1) in more realistic situations with the effect of Hubbard U′ in graphene and hybridization. In realistic graphene zigzag nanoribbon, the Coulomb interaction plays a signification role and stabilizes a spin-polarized ground state30. In this case, the conducting electron at the edge can be regarded as standard one-dimensional Luttinger liquid, coupled with which the two magnetic impurities at low temperature have the RKKY interaction as the asymptotic behavior R−γ cos 2kFR, here kF is the Fermi momentum. The decaying factor γ is the function of Hubbard U′: when U′ = 0, γ = 1, and when U′ ≠ 0, 0 < γ < 1 31,32. So in Luttinger liquid case, at low temperatures the RKKY interaction decays more slowly than the case in Fermi liquid U′ = 0. The methods used to enhance the graphene's SOC in Table 1 are based in the mechanism that via the hybridization between the electronic bands of carbon atoms and heavy atoms with large intrinsic SOC, the SOC of graphene will be greatly enhanced. The hybridization depends on three parameters of heavy atoms: the crystal field, the intrinsic SOC and the hopping between carbon and external atoms. If the crystal field is greatly larger than the later two quantities, the graphene π bands still dominate near the Fermi energy and Kane-Mele Hamiltonian is also valid and the effect of hybridization can be absorbed in the enhanced SOC through second order approximation26 and furthermore when the crystal field decreases and the intrinsic SOC and the hopping between carbon and external atoms increase, the SOC in graphene will increase. While if these three parameters have the same order and crystal field is close to Dirac point, on one hand such strong hybridization can further enhance the SOC in graphene and influence the RKKY interaction as shown in Fig. 2(a)–(c); on the other hand, there are additional bands near the Fermi energy from heavy atoms, the electrons from these bands will exchange with those at the levels of impurities (both spin and charge exchanging). As a result of this exchanging, the spin-spin and charge-charge correlations will increase. In addition, the three ways in Table 1 can raise charge transferring and shift the Fermi energy of pristine graphene from Dirac point, so we do the calculation varying the chemical potential and show the effects in Fig. 2(d).

In conclusion, we investigate two Anderson impurities on a zigzag graphene nanoribbon by QMC simulation in present work. When the system is under a topological phase transition driven by Kane-Mele SOC, we find that, the spin rotation symmetry in RKKY interaction between two impurities can be partly broken because of the SOC. As the two impurities are close to each other and the Fermi energy is near the Dirac point, the in-plane RKKY interactions favor ferromagnetism but the out-plane component can be tuned from ferromagnetism to anti-ferromagnetism by the SOC. The distance effect on RKKY interaction is also systermatically investigated, the in-plane interaction has both behavior of decaying and oscillation, but the out-plane interaction and charge-charge correlation only has decaying as the distance between them is increased. This behavior is totally different from the conventional Fermi liquid. Our theoretical findings here provide some useful suggestions of the spin-polarized STM measurement to RKKY interaction33.

Author Contributions

F.H. conceived the idea and do the QMC calculation, L.K. made the first-principle discussion and collected the data in the Table 1, and F.H. and L.K. wrote the manuscript together. F.H., L.K. and T.F. discussed the results and reviewed the manuscript.

Acknowledgments

This work is supported by DFG Project FOR1616.

References

  1. Castro Neto A. H., Guinea F., Peres N. M. R., Novoselov K. S. & Geim A. K. The electronic properties of graphene. Rev. Mod. P hys. 81, 109–162 (2009). [Google Scholar]
  2. Pesin D. & MacDonald A. H. Spintronics and pseudospintronics in graphene and topological insulators. NatureMater 11, 409–416 (2012). [DOI] [PubMed] [Google Scholar]
  3. Fritz L. & Vojta M. The physics of Kondo impurities in graphene. Rep. Prog. Phys. 76 032501 (2013). [DOI] [PubMed] [Google Scholar]
  4. Uchoa B., Kotov V. N., Peres N. M. R. & Castro Neto A. H. Localized Magnetic States in Graphene. Phys. Rev. Lett. 101, 026805 (2008). [DOI] [PubMed] [Google Scholar]
  5. Hu F. M., Ma T. X., Lin H. Q. & Gubernatis J. E. Magnetic impurities in graphene. Phys. Rev. B 84, 075414 (2011). [Google Scholar]
  6. Wehling T. O., Balatsky A. V., Katsnelson M. I., Lichtenstein A. I. & Rosch A. Orbitally controlled Kondo effect of Co adatoms on graphene. Phys. Rev. B 81, 115427 (2010). [Google Scholar]
  7. Uchoa B., Rappoport T. G. & Castro Neto A. H. Kondo Quantum Criticality of Magnetic Adatoms in Graphene. Phys. Rev. Lett. 106, 016801 (2011). [DOI] [PubMed] [Google Scholar]
  8. Brey L., Fertig H. A. & Sarma S. D. Diluted Graphene Antiferromagnet. Phys. Rev. Lett. 99, 116802 (2007). [DOI] [PubMed] [Google Scholar]
  9. Qi X. L. & Zhang S. C. Topological insulators and superconductors. Rev. Mod. Phys. 83, 1057–1110 (2011). [Google Scholar]
  10. Wu C., Bernevig B. A. & Zhang S. C. Helical Liquid and the Edge of Quantum Spin Hall Systems. Phys. Rev. Lett. 96, 106401 (2006). [DOI] [PubMed] [Google Scholar]
  11. Liu Q., Liu C. X., Xu C. K., Qi X. L. & Zhang S. C. Magnetic Impurities on the Surface of a Topological Insulator. Phys. Rev. Lett. 102, 156603 (2009). [DOI] [PubMed] [Google Scholar]
  12. Zhu J. J., Yao D. X., Zhang S. C. & Chang K. Electrically Controllable Surface Magnetism on the Surface of Topological Insulators. Phys. Rev. Lett. 106, 097201 (2011). [DOI] [PubMed] [Google Scholar]
  13. Maciejko J. et al. Kondo Effect in the Helical Edge Liquid of the Quantum Spin Hall State. Phys. Rev. Lett. 102, 256803 (2009). [DOI] [PubMed] [Google Scholar]
  14. Tanaka Y., Furusaki A. & Matveev K. A. Conductance of a Helical Edge Liquid Coupled to a Magnetic Impurity. Phys. Rev. Lett. 106, 236402 (2011). [DOI] [PubMed] [Google Scholar]
  15. Zhang H. J. et al. Topological insulators in Bi2Se3, Bi2Te3 and Sb2Te3 with a single Dirac cone on the surface. Nature Phys. 5 438–442 (2009). [Google Scholar]
  16. König M. et al. Quantum Spin Hall Insulator State in HgTe Quantum Wells. Science 318 766 (2007). [DOI] [PubMed] [Google Scholar]
  17. Kane C. L. & Mele E. J. Quantum Spin Hall Effect in Graphene. Phys. Rev. Lett. 95, 226801 (2005). [DOI] [PubMed] [Google Scholar]
  18. Hirsch J. E. & Fye R. M. Monte Carlo Method for Magnetic Impurities in Metals. Phys. Rev. Lett. 56, 2521 (1986). [DOI] [PubMed] [Google Scholar]
  19. Hohenadler M. & Assaad F. F. Luttinger liquid physics and spin-flip scattering on helical edges. Phys. Rev. B 85, 081106(R) (2012). [Google Scholar]
  20. Nicolay G., Reinert F. & Hüfner S. Spin-orbit splitting of the L-gap surface state on Au(111) and Ag(111). Phys. Rev. B, 65, 033407 (2001). [Google Scholar]
  21. Dedkov Y. S., Fonin M., Rüdiger U. & Laubscha C. Rashba Effect in the Graphene/Ni(111) System. Phys. Rev. Lett. 100, 107602 (2008). [DOI] [PubMed] [Google Scholar]
  22. Li Z. Y., Yang Z. Q., Qiao S., Hu J. & Wu R. Q. SpinCorbit splitting in graphene on metallic substrates. J. Phys.: Condens. Matter 23, 225502 (2011). [DOI] [PubMed] [Google Scholar]
  23. Marchenko D. et al. Giant Rashba splitting in graphene due to hybridization with gold. Nat. Comm. 3, 12332 (2012). [DOI] [PubMed] [Google Scholar]
  24. Li Y. C. et al. Topological insulators in transition-metal intercalated graphene: The role of d electrons in significantly increasing the spin-orbit gap. Phys. Rev. B 87, 245127 (2013). [Google Scholar]
  25. Abdelouahed S., Ernst A., Henk J., Maznichenko I. V. & Mertig I. Spin-split electronic states in graphene: Effects due to lattice deformation, Rashba effect, and adatoms by first principles. Phys. Rev. B 82, 125424 (2010). [Google Scholar]
  26. Weeks C., Hu J., Alicea J., Franz M. & Wu R. Q. Engineering a Robust Quantum Spin Hall State in Graphene via Adatom Deposition. Phys. Rev. X 1, 021001 (2011). [Google Scholar]
  27. Hu J., Alicea J., Wu R. Q. & Marcel F. Giant Topological Insulator Gap in Graphene with 5d Adatoms. Phys. Rev. Lett. 109, 266801 (2012). [DOI] [PubMed] [Google Scholar]
  28. Kou L. Z. et al. Graphene-Based Topological Insulator with an Intrinsic Bulk Band Gap above Room Temperature. NanoLett. 13, 6251C6255 (2013). [DOI] [PubMed] [Google Scholar]
  29. Kou L. Z. et al. Robust 2D Topological Insulators in van der Waals Heterostructures. ACSNano, 8, 10448 (2014). [DOI] [PubMed] [Google Scholar]
  30. Son Y. W., Cohen M. L. & Louie S. G. Energy Gaps in Graphene Nanoribbons. Phys. Rev. Lett. 97, 216803 (2006). [DOI] [PubMed] [Google Scholar]
  31. Egger R. & Schoeller H. RKKY interaction and Kondo screening cloud for strongly correlated electrons. Phys. Rev. B 54, 16337 (1996). [DOI] [PubMed] [Google Scholar]
  32. Hallberg K. & Egger R. Two-impurity Kondo problem for correlated electrons. Phys. Rev. B 55, R8646 (1997). [Google Scholar]
  33. Zhou L. H. et al. Strength and directionality of surface Ruderman-Kittel-Kasuya-Yosida interaction mapped on the atomic scale. Nat. Phys. 6, 187 (2010). [Google Scholar]

Articles from Scientific Reports are provided here courtesy of Nature Publishing Group

RESOURCES