filling factor reaching a value of g 2.3. A relation to the recent experiment is discussed.
that LL1() is practically depopulated even in an absence However, when the impurity concentration is high
of the impurity, Fig. 5(b)). As a result, the spin den- (ni = 0.2%, two right columns in Fig. 6), the poten-
sity difference, n n , decreases in the impurity region, tials produced by different impurities start to overlap and
which apparently leads to the decrease of P and g in the analysis in terms of a single impurity is no longer
comparison to ideal graphene, see Fig. 5(d). justified. A given value of the filling factor can not be
On the other hand, the influence of the impurity is op- associated with a certain number of the Landau levels,
posite for odd . At = 3 the system is predominantly in since the potential is strongly distorted in comparison to
a unpolarized state, which is manifested by the minimum the ideal case ((a)-panels for ni = 0.2% in Fig. 6) and
of g . However, the distortion of the potential due to the therefore electrons occupy different Landau levels ((b)
impurity gives rise to the formation of a compressible and (c)-panels for ni = 0.2% in Fig. 6). In fact, the de-
strip around the impurity, where EF intersects the LL1. viations in the potential and densities from those of the
This is clearly seen in Fig. 5 (e) where the compressible ideal case become so significant, so the difference between
strip corresponds to regions where the potential is flat the cases of = 2 and = 3 is practically washed out
because of the pinning to EF within the energy window (c.f. two right columns in Fig. 6). As a result, the av-
|E EF | < 2kB T (where 0 < fF D < 1). Because of erage value of the polarization and the effective g-factor
the partial filling of the compressible strip, the electron becomes practically independent of the filling factor.
density there can be easily redistributed there. As a re- In the model used in our calculation the enhancement
sult, the Hubbard interaction pushes up and depopulates of the g-factor is caused by the Hubbard term in the po-
the LL1() while the LL1() remains populated, see Figs. tential, Eqs. (6) and (9). Let us briefly discuss how the
5(f),(g). This leads to a local spin polarization around calculated value of g depends on the Hubbard constant
the impurity as illustrated in Fig.(5)(h). therefore, the U . While we used value U 3.5t29 , the current litera-
overall polarization is no longer zero, hP (r)i > 0 and, ture reports various estimations of U in the range 0.5t .
hence, the effective g-factor does not drop to the mini- U . 2t,3335 where t 2.7 eV is the hopping integral in
mum value, remaining g > gmin
. the standard p-orbital tight-binding Hamiltonian1 . We
Summarizing, the influence of a single impurity calculated the dependencies g = g () for different val-
is twofold: when the system is predominantly spin- ues of the parameter U and found that the results show
polarized (even ), the impurity decreases the average the same qualitative behaviour and the calculated value
polarization and the effective g-factor by locally pushing- of g scales linearly with U. This is illustrated in the inset
up the Landau levels and depopulating them; in the op- to Fig. 2 which shows a dependence of the effective g-
posite case of a predominantly non-polarized system (odd factor on the Hubbard constant for a representative value
), the impurity leads to the local formation of the spin- of = 2.5.
polarized compressible strips, which instead increases the Let us now discuss the relation of our findings to the
average polarization and the effective g-factor. recent experiment. Measurements done by Kurganova
Having understood the effect of a single impurity on et al.20 exhibit the enhancement of the effective spin-
the average polarization and the effective g-factor it is splitting leading to the effective g-factor g = 2.7 0.2.
straightforward to generalize the obtained results for an Also, the enhanced effective g-factor was found to be
arbitrary concentration of impurities. The higher con- practically independent on . Our calculations show that
centration ni , the larger influence of impurities on the for low impurity concentrations, g exhibits a pronounced
average value of the spin-polarization and the effective oscillatory behavior in the range 2 g . 4, and it
g-factor. As a result, an increase of the impurity con- becomes rather independent of for larger ni reaching
centration leads to the suppression of the amplitude of a saturated value g 2.3. Our calculations therefore
oscillations as shown in Fig. 2. strongly suggest that impurities always present in realis-
Note that for a sufficiently large impurity concentra- tic samples play an essential role in suppressing the os-
tion (in our case ni = 0.2%), the oscillations of g get cillatory behavior of g . Note that in real systems the
practically suppressed and g becomes rather indepen- oscillations of g can be smoothed by a number of addi-
dent on the filling factor, see Fig. 2. This effect can be tional factors. The measurements of Kurganova et. al 20
understood from a comparison of two distinct cases of were performed in a tilted magnetic fields and at large
low and high impurity concentration, ni = 0.02% and filling factors > 6. In this case the distance between
ni = 0.2%, see Fig. 6. When the impurity concentration the adjacent Landau levels is comparable to the Zeeman
is low (ni = 0.02%, two left columns in Fig. 6), the self- splitting which results in stronger overlap of the succes-
consistent potentials produced by different impurities do sive Landau levels and eventually leads to an additional
not overlap and the system can be treated as an assembly smearing of g . Therefore our calculations motivate for
of independent impurities. (The potential is flat every- further studies of the effective g-factor close to = 0,
where besides narrow regions close to the impurities, see where the oscillatory behavior of g is expected to be
(a)-panels for ni = 0.02% in Fig. 6). At = 2 the pres- more pronounced. Our finding also indicate that the os-
ence of impurities decreases locally polarization (dips on cillatory behavior of the effective g-factor is expected to
(c)-panel), while at = 3 the local polarization increases be more pronounced in suspended samples where the in-
(peaks on (c)-panel). fluence of charged impurities will be much less important.
6
FIG. 6. (Color online). The spin resolved potential, densities and polarization for different concentration charged of
impurities(ni = 0.2%, 0.02%) and for different filling factors ( = 2, 3). The one-dimensional plots of (a) V (y), (b) the
spin-up and (c) spin-down charge densities, and (d) the spin polarization P (y) as a functions of y for x = 50 nm. Dashed lines
correspond to the ideal system (without impurities). (e) The 2D plot of the spatially resolved spin polarization P (x, y) in a
graphene sheet. The system parameters are Nx = 800, Ny = 461, d = 10 nm, B = 50 T.
Finally, it is noteworthy that spin-splitting of in showed that for low impurity concentration g oscillates
graphene36 and graphene quantum dots37 was also ex- as a function of the filling factor in the range from
perimentally studied in a parallel magnetic filed. It was gmin = 2 to gmax 4 reaching maxima at even filling
concluded that in this case the effective g-factor does dif- factors and minima at odd ones. Finally, we outlined the
fer from its free-electron value. This can be explained by influence of impurities on the spin-splitting and demon-
the fact that in the parallel field the Landau levels do not strated that the increase of the impurity concentration
form and therefore the interaction induced enhancement leads to the suppression of the oscillation amplitude and
of the g -factor is small. to a saturation of the the effective g-factor around a value
of g 2.3.
IV. CONCLUSIONS
ACKNOWLEDGMENTS
In this work we employed the Thomas-Fermi approxi-
mation in order to study the effective g-factor in graphene
in the presence of a perpendicular magnetic field taking We acknowledge a support of the Swedish Research
into account the effect of charged impurities in the sub- Council (VR) and the Swedish Institute (SI). A.V.V. also
strate. We found that electron-electron interaction leads acknowledges the Dynasty foundation for financial sup-
to the enhancement of the spin splitting, which is char- port. The authors are grateful to V. Gusynin for critical
acterized by the increase of the effective g-factor. We reading of the manuscript.
2
on ton@mail.ru K. S. Novoselov, Z. Jiang, Y. Zhang, S. V. Morozov, H. L.
4 21
T. Ando and Y. Uemura, J. Phys. Soc. Jpn. 37, 1044 S. Ihnatsenka and I. V. Zozoulenko, arXiv:1206.6251v1
(1974). [cond-mat.mes-hall] (2012).
5 22
L. Zhang, Y. Zhang, M. Khodas, T. Valla, and I. A. Zal- Enrico Rossi and S. Das Sarma, Phys. Rev. Lett. 101,
iznyak, Phys. Rev. Lett. 105, 046804 (2010). 166803 (2008).
6 23
T. Englert, D. Tsui, A. Gossard, and C. Uihlein, Surf. Sci. K. Lier and R. R. Gerhardts, Phys. Rev. B 50, 7757 (1994).
24
113, 295 (1982). J. H. Oh and R. R. Gerhardts, Phys. Rev. B 56, 13519
7
J. M. Kinaret and P. A. Lee, Phys. Rev. B 42, 11768 (1997).
25
(1990). W.-R. Hannes, M. Jonson and M. Titov, Phys. Rev. B 84,
8
J. Dempsey, B. Y. Gelfand, and B. I. Halperin, Phys. Rev. 045414 (2011).
26
Lett. 70, 3639 (1993). A. A. Shylau, J. W. Klos, and I. V. Zozoulenko, Phys. Rev.
9
Y. Tokura and S. Tarucha, Phys. Rev. B 50, 10981 (1994). B 80, 205402 (2009).
10 27
Z. Zhang and P. Vasilopoulos, Phys. Rev. B 66, 205322 J. Fernandez-Rossier, J. J. Palacios, and L. Brey, Phys.
(2002). Rev. B 75, 205441 (2007).
11 28
T. H. Stoof and G. E. W. Bauer, Phys. Rev. B 52, 12143 To facilitate calculations of V (r) the summation in Eq.(5)
(1995). was performed numerically in the region |r r | < R. Out-
12
S. Ihnatsenka and I. V. Zozoulenko, Phys. Rev. B 73, side this region the charge density was assumed to be uni-
075331 (2006). form n(r) = hni, such that that the additional potential
13
S. Ihnatsenka and I. V. Zozoulenko, Phys. Rev. B 73, produced by hni by charges in the this region was evaluated
155314 (2006). analytically.
14 29
S. Ihnatsenka and I. V. Zozoulenko, Phys. Rev. B 74, T. O. Wehling, E. Sasioglu, C. Friedrich, A. I. Lichten-
075320 (2006). stein, M. I. Katsnelson and S. Blugel, Phys. Rev. Lett.
15
S. Ihnatsenka and I. V. Zozoulenko, Phys. Rev. B 78, 106, 236805 (2011).
30
035340 (2008). A. A. Shylau and I. V. Zozoulenko, Phys. Rev. B 84,
16
A. Ghosh, C. J. B. Ford, M. Pepper, H. E. Beere, and D. 075407 (2011).
31
A. Ritchie, Phys. Rev. Lett. 92, 116601 (2004). J. H. Davies, The physics of low-dimensional semiconduc-
17
A. R. Goni, P. Giudici, F. A. Reboredo, C. R. Proetto, tors: an introduction, (Cambridge university press, Cam-
C. Thomsen, K. Eberl, and M. Hauser, Phys. Rev. B 70, bridge, 1998).
32
195331 (2004). D. B. Chklovskii, B. I. Shklovskii, and L. I. Glazman, Phys.
18
M. Evaldsson, S. Ihnatsenka, and I. V. Zozoulenko, Phys. Rev. B 46, 4026 (1992).
33
Rev. B 77, 165306 (2008). O. V. Yazyev, Phys. Rev. Lett. 101, 037203 (2008).
19 34
J. M. Schneider, N. A. Goncharuk, P. Vasek, P. Svoboda, J. Jung and A. H. MacDonald, Phys. Rev. B 80, 235417
Z. Vyborny, L. Smrcka, M. Orlita, M. Potemski, and D. (2009).
35
K. Maude, Phys. Rev. B 81, 195204 (2010). C. Tao, L. Jiao, O. V. Yazyev, Y.-C. Chen, J. Feng, X.
20
E. V. Kurganova, H. J. van Elferen, A. McCollam, L. A. Zhang, R. B. Capaz, J. M. Tour, A. Zettl, S. G. Louie, H.
Ponomarenko, K. S. Novoselov, A. Veligura, B. J. van Dai, and M. F. Crommie, Nat. Phys. 7, 616 (2011).
36
Wees, J. C. Maan, and U. Zeitler, Phys. Rev. B, 84, Mark B. Lundeberg and Joshua A. Folk, Nat. Phys. 5, 894
121407, (2011). (2009)
37
J. Guttinger, T. Frey, C. Stampfer, T. Ihn, and K. Ensslin,
Phys. Rev. Lett. 105, 116801 (2010).