You are on page 1of 6

ISSN 1063-7826, Semiconductors, 2018, Vol. 52, No. 3, pp. 335–340. © Pleiades Publishing, Ltd., 2018.

Original Russian Text © S.Yu. Davydov, 2018, published in Fizika i Tekhnika Poluprovodnikov, 2018, Vol. 52, No. 3, pp. 353–358.

CARBON
SYSTEMS

Electron–Electron and Electron–Phonon Interactions in Graphene


on a Semiconductor Substrate: Simple Estimations
S. Yu. Davydova, b*
a Ioffe
Institute, St. Petersburg, 194021 Russia
b St. Petersburg National Research University of Information Technologies, Mechanics, and Optics,
St. Petersburg, 197101 Russia
*e-mail: Sergei_Davydov@mail.ru
Submitted May 25, 2017; accepted for publication June 22, 2017

Abstract—The problem of epitaxial graphene formed on a semiconductor substrate is considered in the con-
text of the extended Hubbard and Holstein–Hubbard models for electron–electron and electron–phonon
interactions. The Haldane–Anderson model is chosen for the density of states of the substrate. Three regions
of the phase diagram, specifically, spin- and charge-density waves and a spin- and charge-homogeneous
paramagnetic state are considered. For a number of special cases used as examples, the similarities and dif-
ferences of the electron states of graphene on semiconductor and metal substrates are demonstrated. It is
shown that the main difference arises, if the Dirac point of graphene lies within the band gap of the semicon-
ductor. Numerical estimations are performed for a SiC substrate.

DOI: 10.1134/S1063782618030090

1. INTRODUCTION and the density of states (DoS) is never zero. This is


The influence of Coulomb repulsion on the elec- because of the fact that the DoS corresponding to the
tronic spectrum of free graphene (FG) is the subject of conduction band of the metal is finite in the entire
a considerable number of studies (see, e.g., [1–8]). In energy region.
studies, a popular model is the standard Hubbard In this study, the EHM [9] is applied to the
model (HM) which takes into account the intra- description of single-layer graphene on a semiconduc-
atomic correlation of electrons U (one-center repul- tor. We here use also the results obtained in [10], where
sion). The extended version of the HM, i.e., the the electronic structure of graphene-like compounds
extended Hubbard model (EHM) which additionally on a semiconductor substrate was studied. The point is
includes the interatomic interaction of electrons G that, in the general case, consideration for electron
(intercenter repulsion) is used as well. In the context of correlations in graphene brings about a difference
the HM and EHM, it is possible to determine the between the occupation numbers nA and nB of sublat-
regions of parameters that provide the homogeneous tices A and B of graphene. The same situation (nA ≠ nB)
Dirac semimetal (SM) state, in which each carbon takes place in graphene-like ANB8 – N binary com-
atom corresponds to one electron with zero total spin
pounds as well, since the sublattices A and B are occu-
moment, and states with inhomogeneous spin and
pied by chemically different atoms. It is worth noting
charge distributions of electrons among the sublat-
that the inequality nA ≠ nB corresponds to the presence
tices, i.e., spin-density waves (SDWs) and charge-
density waves (CDWs). In contrast to the Dirac SM of a gap in the electronic spectrum and the DoS. Thus,
state exhibiting a zero-gap electronic spectrum, the graphene in the SDW or CDW state is formally equiv-
SDW and CDW states in FG exhibit a gap and are alent to a graphene-like compound.
Mott insulators. In studying the influence of electron–phonon
In [9], the EHM was used to describe epitaxial interaction, it is convenient to use the Holstein–Hub-
graphene (EG) on a metal substrate and to determine bard model (HHM) which takes into account also
the boundaries of the SM, SDW, and CDW regions. electron–electron repulsion (as a rule, one-center
Among other things, it was shown that, in contrast to repulsion) [11–20]. The HHM is applied to the
the spectrum of FG, the spectrum of EG on metal in description of superconductivity (including high-tem-
the SDW and CDW states does not involve gaps (i.e., perature superconductivity), SDWs and CDWs, and
regions of forbidden states), the spectrum of EG in the polarons and bipolarons in three-, two- and one-
SM state in the vicinity of the Dirac point is nonlinear, dimensional (3D, 2D, and 1D) lattices, and clusters.

335
336 DAVYDOV

In [21], a model of EG on metal is proposed. The these states are formed not only due to the kinetic
model combines the EHM and the Holstein model exchange, or t interaction (see below), but due to elec-
(extended Holstein–Hubbard model (EHHM)) tron correlations and electron–phonon coupling as
which describes the interaction of a band electron with well.
an Einstein phonon. As in [9], three regions of the
phase diagram, i.e., the SDWs, CDWs, and the state
homogeneous in spin and charge are considered. Here 2. ELECTRON–ELECTRON REPULSION
we extend the results of [21] to the case of a semicon-
ductor substrate. 2.1. General Relations
Thus, the purpose of this study is to clarify the In [9], it is shown that, in the Hartree–Fock
question of how the band gap in the electronic spec- approximation, the diagonal Green’s functions of EG
trum of the substrate influences the states of EG, if for the sublattices A and B in the EHM are

σ ΩσB( A) + i Γ(ω)
G AA (BB )(ω, k) = , (1)
(ΩσA + i Γ(ω))(ΩσB + i Γ(ω)) − t 2 f 2(k)
f (k) = 3 + 2cos(k x a0 3) + 4cos(k x a0 3/2)cos(3k y a0 /2). (2)

Here, k = (kx, ky) is the wave vector for electron sion for the DoS of an EG planar sheet on a semicon-
motion in the graphene-sheet plane; a0 are the spacing ductor as
between nearest neighbors in graphene, ω is the
Γ(ω) |ξ + Bσξ + Cσ |
4 2
energy variable; t is the energy of electron transfer ρσ (ω) =
EG
ln
between nearest neighbors in graphene (kinetic πξ2 Cσ
exchange); (4)
σ⎛
2Ω 2ξ + Bσ
2
Bσ ⎞
+ 2 ⎜
− arctan
 σ ⎟⎠
arctan .
ΩσA(B ) = ω − εσ(ω) ∓ Δ σ(ω), 2εσ(ω) = (ε Aσ + ε Bσ ), πξ ⎝ 4Γ(ω)Ωσ 4Γ(ω)Ω
Here, Ω  σ = ω − ε σ , 2ε σ = (ε Aσ + ε Bσ ), 2Λσ =
2Δ σ(ω) = (ε Aσ + ε Bσ ), ε A(B )σ = ε A(B )σ + Λ(ω),
(ε Aσ − ε Bσ ) , ξ = t 2π 3, Gσ = −2(Ω  2σ − Δ 2σ − Γ2sc(ω)),
where σ is the spin-indicating superscript and sub-
script; and Cσ = (Ω  2σ − Δ 2σ )2 + Γ2sc(ω)(Γ2sc(ω) + 2Δ 2σ + 2Ω
 2σ ).
The energy levels εAσ and εBσ of atoms of the sublat-
Γ(ω) = πV 2ρsub (ω), tices are determined as
and ε A ↑,↓ = ε0 + Λ(ω) + U (a ∓ b + c ∓ d )/2 + zG (a − c),

ε B ↑,↓ = ε0 + Λ(ω) + U (a ∓ b − c ∓ d )/2 + zG (a + c). (5)
Λ(ω) = V 2P ∫
−∞
d ω 'ρsub (ω')/(ω − ω'),
Here, z = 3 is the number of nearest neighbors in
graphene and the parameters a, b, c, and d are related
where ρsub(ω) is the DoS of the substrate, are the to the occupation numbers of atoms of the sublattices
broadening and shift functions, respectively; V is the nAσ and nBσ by the expressions
matrix element of graphene–substrate interaction;
and P is the symbol of the principal value. 2a = nA ↑ + nA ↓ + nB ↑ + nB ↓,
The electronic spectrum of the system is deter- 2b = nA ↑ − nA ↓ + nB ↑ − nB ↓,
mined from the equation ΩσAΩσB = t 2 f 2(k) (the real 2c = nA ↑ + nA ↓ − nB ↑ − nB ↓,
(6)
poles of Green’s function (1)), which gives 2d = nA ↑ − nA ↓ − nB ↑ + nB ↓,
E ±σ(ω, k) = εa (ω) ± Rσ(ω, k), so that we have
nA ↑,↓ = (a ± b + c ± d )/2, nB ↑,↓ = (a ± b − c ± d )/2.
Rσ(ω, k) = Δσ(ω) + t f (k).
2 2 2
(3)
The energy of the system is
In (3), the minus sign corresponds to the valence π
band, and the plus sign to the π* conduction band. E = ∑ε
α,σ
ˆασ
ασ n −U ∑n α
α↑ nα↓ − zGnA nB
In accordance with the results of [9, 10], where the εF −εασ (7)
function f(k) was described in the low-energy approx-
imation f(k) ≈ 2a0|k|/2 [1], we write the general expres-
+ ∑ ∫
α,σ
σ
Ωσρα (Ωσ )d Ωσ + Esub,
−∞

SEMICONDUCTORS Vol. 52 No. 3 2018


ELECTRON–ELECTRON AND ELECTRON–PHONON INTERACTIONS 337

where εF is the Fermi level, ρσα (Ωσ ) is the DoS for the Now, we pass to EG. Figure 1 in [10] schematically
sublattice α = (A, B) (per atom), and Esub is the contri- illustrates different cases of mutual arrangement of
bution of the substrate to the energy of the system. energy bands upon contact between a graphene-like
To specify the broadening function Γ(ω) and the compound and a semiconductor substrate with no
shift function Λ(ω), it is necessary to preset the DoS of interaction between them. These energy diagrams are
the semiconductor substrate ρsub(ω). To do this, we analogues of common semiconductor heterojunctions
choose the Haldane–Anderson model (see, e.g., [24].
[22, 23]). In accordance with this model, we have We begin from the case of a broken heterojunction
(Fig. 1d in [10], heterojunction of type III). It is clear
⎧ρsc, |ω| > E g /2, that, in the limit of |εσ| ≫ Eg, we essentially have the
ρsub (ω) = ⎨ (8)
⎩0, |ω| ≤ E g /2, case of EG on metal, which was considered in detail in
[9]. In fact, according to (8), the DoS of the semicon-
where ρsc = const and zero energy corresponds to the ductor substrate in the above-indicated energy region
middle of the band gap Eg. Then the broadening func- is a constant, as in the case of a metal substrate [9]. It
tion is Γ(ω) = Γsc = πV2ρsc at |ω| > Eg/2 and Γ(ω) = 0 can be easily shown that, in this limit, there is notice-
at |ω| ≤ Eg/2. The shift function Λ(ω) is able charge transfer. In fact, at εσ < –Eg, an electron
transitions from the valence band of the substrate to
Γsc ω − E g /2 the conduction band of EG. At εσ > Eg, the process is
Λ(ω) = ln . (9)
π ω + E g /2 the reverse: an electron from the valence band of EG
The set of formulas (1)–(9) completely defines the tunnels to the conduction band of the substrate. Thus,
problem formulated above. The densities of states as in [9], the doping of EG results in widening of the
expressed by formula (4) were considered at different region of the SM state, and the effect is independent of
relations between the parameters of the problem pre- the sign of free charge carriers introduced into EG by
viously (see [9, 10, 22, 23] and references therein). the substrate.
Similarly to what was done in [9], we consider a num- Let us consider this case more thoroughly. Since
ber of special cases for illustration. |Λ(ω)| ∝ |ω|–1, we assume Λ(ω) = 0 and rewrite the set
of Eqs. (27) from [9] in the form
2.2. Graphene–Semiconductor Heterojunctions −Γsc tan[π(ν + c + d )/2] = δ + U (c − d )/2 − czG,
At first, we consider FG assuming V = 0. In this −Γsc tan[π(ν + c − d )/2] = δ + U (c + d )/2 − czG, (11)
case, the functions Γ(ω) and Λ(ω) are identically zero,
and we obtain the DoS of FG as where δ = ε* – εF. We assume |δ| ≫ Eg, a = 1 + ν, |ν| ~ 1.
At δ > 0, we have ν ~ –1, and at δ < 0, we have ν ~ 1.
⎧2|Ωσ | . R ≥ |Ω | ≥ |Δ |, If in both equations (11) the right-hand sides are pos-
⎪ σ σ σ
itive and large, then we have ν + c + d ~ –1 and ν +
ρFG
σ (Ωσ ) = ⎨ ξ2 (10)
⎪⎩0, |Ωσ | < |Δσ |, |Ωσ | > Rσ. c – d ~ –1. Similarly, if the right-hand sides of equa-
tions (11) are negative and large in absolute value, then
Here, Ωσ = ω − εσ , 2εσ = (ε Aσ + εBσ ), 2Δσ = ν + c + d ~ 1 and ν + c – d ~ 1. Hence it follows that
c = d = 0; i.e., in EG, the analogue of the SM state of
(ε Aσ − εBσ ) , Rσ = ξ2 + Δ2σ , and εA(B)σ = ε A(B )σ at FG is formed. However, this state is not the Dirac SM
Λ(ω) = 0 (see [9]). As in [9], we assume ε* ≡ state that features a zero DoS at the Dirac point and a
ε0 + a(U /2 + zG ) = εF; then we obtain a = 1 and nA,B = linear energy dependence of the density of states near
1 ± c. If the total magnetization of the unit cell of the Dirac point. In the case under consideration, the
graphene is zero (i.e., b = 0), for the spin momenta DoS at the Dirac point is finite, and the energy
mA,B = nA ↑,B ↑ − nA ↓,B ↓ , we obtain the equality mA = dependence of the DoS in the vicinity of the Dirac
point is nonlinear. Thus, in the zero approximation
–mB = d corresponding to the antiferromagnetic state. in the limit of |εσ| ≫ Eg, the difference between the
In this case, the phase diagram of FG involves three cases of a metal substrate and a semiconductor sub-
regions [9]: the zero-gap SM state, which is homoge- strate is reduced to the replacement of Γm with Γsc.
neous in charge and zero spin and described by the
Now we pass to straddling heterojunctions of type I
dispersion relation E ±σ (k) = ±t |f (k)| ; the SDW state (Figs. 1a, 1b in [9]). In EG, in the energy range of the
corresponding to the dispersion relation E ±σ (k) = band gap, |ω| ≤ Eg/2, the function Γ(ω) is identically
zero, whereas the function Λ(ω) remains finite and
± (dU /2)2 + t 2 f 2(k); and the CDW state with the dis- equal to zero only at ω = 0 (see (9)). Thus, the first
persion relation E ±σ (k) = ± c 2(U /2 − zG )2 + t 2 f 2(k). term in (4) is vanishing. The second term in (4) is van-
It can be easily seen that the boundary between the ishing as well, if the numerators of the arguments of
SDW and CDS states is defined by the condition arctangent functions are identical in sign, i.e., satisfy
zG = U. the inequality Bσ(2ξ2 + Bσ) > 0. In this case, the DoS

SEMICONDUCTORS Vol. 52 No. 3 2018


338 DAVYDOV

of EG is identically zero. The above inequality is satis- ment of the covalent bonding of a carbon atom of
fied at graphene with an atom of the substrate is V =
 σ | < |Δσ |, |Ω
|Ω  σ | > Rσ, ηb (2 /m0a⊥2 ), where a⊥ is the graphene–substrate
spacing, m0 is the free electron mass, ℏ is Planck’s
where constant, and ηb is the numerical coefficient for the σ
Ω σ = ω − εσ − Λ sc(ω). bond. The coefficient ηb depends on the type of
In the region of energies that satisfy the condition hybridized orbitals forming the corresponding σ bond.
 σ | ≤ Rσ , the DoS (formula (4)) coincides with For sp3 orbitals, we have ηb = 3.22 [26, 27]. For the
|Δσ| ≤ |Ω
σ bond of the pz orbital of an atom of graphene with an
expression (10), if Ωσ = ω – εσ is replaced with Ω σ =
sp3 orbital of the substrate, we obtain ηb = 2.63 [28]. As
Ωα − Λ(ω). In the limit of |ω| ≪ Eg, the shift function in [21], we assume w⊥ = −∂V /∂a⊥ and then obtain
is Λ(ω) ≈ −ωΓsc /πE g . Disregarding the shift Λ(ω) in w⊥ = 2V /a⊥ . According to calculations performed in
the zero approximation, we arrive at the case of FG. [28], we take a⊥ = 2 Å and, thus, obtain V ≈ 6 eV and
The type-II staggered heterojunction (Fig. 1c w⊥ ≈ 3 eV Å–1 in the former case and V ≈ 5 eV and w⊥ ≈
in [9]) is intermediate with respect to heterojunctions 2.5 eV Å–1 in the latter case. However, the experimen-
of types I and III and is not considered here separately.
Thus, with some simplification, we can state that, in tal data [30] show a⊥ ≈ 3 Å–1. This spacing corresponds
the case of a semiconductor substrate, the DoS of EG to van der Waals interaction rather than covalent inter-
can vary from the DoS of FG to the DoS of EG on a action. Using the results of [21], we obtain w⊥ =
metal substrate, depending on the parameters of the 12D/a⊥, where D is the energy of the van der Waals
problem. More detailed consideration of heterojunc- bonding of graphene with the substrate. In the case of
tions in the context of the cluster model [9, 10] is given covalent interaction, it is appropriate to refer to
in the Appendix. Some numerical estimates of the graphene as a buffer layer, and in the case of van der
parameters are given in the Appendix as well. Waals interaction, as quasi-free graphene [21].
Now we estimate the central force constant k0⊥ =
3. ELECTRON–PHONON INTERACTION M ω2⊥ [21, 31]. We make use of the results of [32] and,
Analysis of the role of electron–phonon interac- for simplicity of estimations, disregard the ionicity and
tion in the context of the EHHM [21] shows that, metallicity of bonding. Then for covalent bonding, we
instead of expression (5) for the energy levels εAσ and write k0⊥ ≈ 4V /a⊥2 and, hence, obtain k0⊥ ≈ 6 eV Å–2 for
εBσ, we obtain the σ bond formed by the sp3 orbitals of atoms of
ε A ↑,↓ = ε0 + Λ(ω) + U (a ∓ b + c ∓ d )/2 graphene and the substrate and k0⊥ ≈ 5 eV Å–2 for the
σ bond formed by the pz orbital of an atom of graphene
+ zG (a − c) − λ(a + c),
and the sp3 orbital of the substrate. For van der Waals
ε B ↑,↓ = ε0 + Λ(ω) + U (a ∓ b − c ∓ d )/2 interaction, we have k0⊥ = 72D/a⊥2 [21]. Hence, it fol-
(12)
+ zG (a + c) − λ(a − c).
lows that we have λ⊥ ≈ w⊥2 /k0 ⊥ , so that λ = 1.50 and
Here, the overall constant of electron–phonon inter- 1.25 eV for covalent bonding of the sp3 and pz orbitals
action is λ = λ|| + λ⊥, where λ|| and λ⊥ are the constants of graphene with the sp3 orbital of the substrate, and
of electron–phonon interaction for the vibrations of λ⊥ = 2D in the case of van der Waals interaction.
carbon atoms in the graphene-sheet plane and for In order of magnitude, the energy D is several tenths
vibrations of the graphene sheet with respect to the of an eV.
substrate, respectively.
As shown in [21], the constant of electron–phonon
coupling for a long-wavelength optical phonon with 4. DISCUSSION
the energy ω|| in FG is λ|| = w||2 /M ω||2 , where M is the Thus, the application of EHM to the case of a
carbon atomic mass and w|| = 2t/a0 is the deformation semiconductor substrate shows that, in general, the
constant. Taking t = 2.8 eV and ω|| ≈ 0.2 eV, we obtain phase diagram of EG is qualitatively the same as that
w|| ≈ 3.9 eV Å–1 and λ|| ≈ 0.13 eV [21]. in the case of a metal substrate [9]. The model param-
Let us now estimate the deformation constant w⊥ eters for these substrates are naturally different, result-
and the constant of electron–phonon coupling λ⊥ = ing in a shift of the boundaries between regions of the
phase diagram. The same can be said about the use of
w⊥2 /M ω2⊥ for vibrations of the graphene sheet (as a the EHHM in the case of a semiconductor substrate.
whole) with respect to a semiconductor substrate at For example, it can be easily shown that, at a = 1 and
the frequency ω⊥. In accordance with the Harrison b = 0, the role of electron–phonon interaction is
method of bonding orbitals [25–27], the matrix ele- reduced to renormalization of the intercenter repul-

SEMICONDUCTORS Vol. 52 No. 3 2018


ELECTRON–ELECTRON AND ELECTRON–PHONON INTERACTIONS 339

sion G which is replaced with G* = G + λ/z. Since U ≈ and in the region of the band gap, |ω| < Eg, we have
9.3 eV and G ≈ 5.5 eV [33], for the most interesting case
of quasi-free graphene, we obtain U, G ≫ λ/z, i.e., the ρ σ± (ω) = δ(ω – ε ±σ ), where δ(…) is the Dirac δ func-
same result as in the case of graphene on metal [21]. tion. At first, we consider the case of ε* = εF = 0 (a = 1
Moreover, for strong graphene–substrate coupling, and b = 0) and |εσ| ≫ Eg. To determine the occupation
we have λ⊥ ≈ 1 eV (metal) and 1.25–1.50 eV (semicon- numbers, we resort to additional simplification often
ductor). Thus, from the standpoint of parameters of used in the theory of adsorption [23] and replace Λ(ω)
the problem, the differences are small, and all of the and Γ(ω) with Λ(ε±σ) and Γ(ε±σ). Then we have
conclusions related to the diagrams of states of EG Γ(ε±σ) = Γsc. Taking into account the fact that, at the
[9, 21] are valid also in the case considered here. It is energies |ω| ≫ Eg, the shift function is Λ(ω) ≈
worth noting that, in the experiment [34] with EG on −ΓscE g /πω , we obtain ε ±σ ≈ ±tσ* , where tσ* =
Ir and SiC substrates, no radical differences were
noticed as well. tσ(1 − ΓscE g /πztσ2 ) . As in [22], we assume that, for
However, from the standpoint of the theory devel- covalent graphene–substrate coupling, Veff = V/3 ~
oped in this study, such differences can occur, if the 2 eV and ρsub = 0.25 eV–1, and then obtain Γsc ≈ 3 eV.
Dirac point of EG falls within the band gap of the We take into account that t ~ 3 eV [1] and Eg ~ 3 eV (the
semiconductor substrate. If CDW or SDW states are 4H, 6H, 8H, 15R, 18R, and 21R silicon-carbide poly-
formed in this case (which is possible in the case of types [35]). Then we have ΓscEg(πzt2)–1 ~ (zπ)–1 ~ 0.1.
quasi-free EG [9]), the spectrum and the DoS of EG In the case of van der Waals bonding, Γsc ~ D, so that
exhibit band gaps, as in hexagonal graphene-like lay- ΓscEg(πzt2)–1 ≪ 1. Thus, all of the manipulations are
ers (see Figs. 2, 3 in [10]). In the case of a metal sub-
strate, such band gaps are impossible. In conclu- reduced to renormalization of the parameter tσ. Since
sion, it should be noted that the EHM and EHHM tσ ≥ zt, the correction can always thought to be small.
are applicable also to EG on insulator substrates, In the case under consideration, the set of Eqs. (26)
such as, e.g., SiO2. from [9] transforms into the set

APPENDIX π(c − d ) = 2C↓* arctan(t↓* /Γsc ),


To simplify the problem, we consider a six-atom
cluster model that proved to be good in [9] (see Fig. 3). π(c + d ) = 2C↑* arctan(t↑* /Γsc ), (A.6)
Extending the results of [9] to the case of a semicon-
ductor substrate, we write the diagonal Green’s func- where Cσ* = |Δσ |/tσ* and Δ↑,↓ = U (c ∓ d )/2 − czG .
tions for atoms of the sublattices A and B as Extending the results of [9] to the case under consid-
eration, we can argue that the solution c ≠ 0, D = 0
D AA
σ
(ω) = 1 (1 − Cσ )g+σ(ω) + 1 (1 + Cσ )g−σ(ω), (inhomogeneous charge-density (ICD) distribution)
2 2 is valid, if the inequality 2zG – U > πΓscη*, where
D BB (ω) = (1 + Cσ )g+ (ω) + (1 − Cσ )g−σ(ω). (A.1)
σ 1 σ 1 η* = τ arctan(τ) with τ = (t* z /Γsc ) and t* =
2 2
t(1 − ΓscE g /πzt 2 ) is satisfied. The solution c = 0 and
Here,
d ≠ 0 (inhomogeneous spin-density (ISD) distribution)
g±σ(ω) = (ω − ε ±σ + i Γ(ω)), (A.2) is formed at the condition U > πΓscη*. The boundary
between the ICD and ISD states corresponds to the con-
and dition c = d ≠ 0 and the inequality U = zG > πΓscηmax,
ε ±σ = εσ + Λ(ω) ± tσ, tσ = Δ2σ + zt 2, where ηmax = (t * /Γ )/ arctan(t * /Γ ) and
max sc max sc
(A.3)
Cσ = |Δ σ |/tσ.
* = (zG )2 + z(t*)2 {1 − Γ E /π[(zG )2 + z(t*)2 ]}.
tmax
The corresponding DoSs are sc g

ρ σAA (ω) = 1 (1 − Cσ )ρ σ+ (ω) + 1 (1 + Cσ )ρ σ− (ω), It can be easily seen that, if we assume Eg = 0, we
2 2 obtain the results of [9] for a metal substrate. The cor-
responding energy diagram is shown in Fig. 4 [9], in
ρ σBB (ω) = 1 (1 + Cσ )ρ σ+ (ω) + 1 (1 − Cσ )ρ σ− (ω). (A.4) which Γm and η should be replaced with Γsc and η*.
2 2
At the same time, if we assume t = 0, we arrive at a spe-
In the region of allowed states, |ω| > Eg, the DoSs cial case, for which the phase diagram is shown in
ρ σ± (ω) are Fig. 2 in [9], with the replacement of Γm with Γsc.
Finally, at Γsc → 0, we arrive at the case of FG, with
Γsc
ρ ± (ω) = 1
σ consideration for interatomic t interaction, however,
, (A.5)
π (ω − ε ±σ )2 + Γ2sc in the cluster version rather than in the band version.

SEMICONDUCTORS Vol. 52 No. 3 2018


340 DAVYDOV

In this case, for the ISD state to exist, the inequality 19. I. V. Sankar and A. Chatterjee, Physica B 489, 17
(2016).
U > 2 zt* ≈ 3.5t* should be satisfied, and for the ICD
20. T. Miyao, arXiv: 1402.5202; arXiv: 1610.09039.
state to exist, the inequality 2zG – U > 2 zt* should be 21. S. Yu. Davydov, Semiconductors 52 (2018, in press).
satisfied. 22. S. Yu. Davydov, Semiconductors 47, 95 (2013).
23. S. Yu. Davydov, The Theory of Adsorption: Method of
REFERENCES Model Hamiltonians (SPbGETU LETI, and St. Peters-
burg, 2013) [in Russian]. twirpx.com/file/1596114/.
1. A. H. Castro Neto, F. Guinea, N. M. R. Peres, twirpx.com/file/1014608/.
K. S. Novoselov, and A. K. Geim, Rev. Mod. Phys. 81,
109 (2009). 24. T. Bechshtedt and R. Enderlein, Semiconductor Sur-
faces and Interfaces. Their Atomic and Electronic Struc-
2. V. N. Kotov, B. Uchoa, V. M. Pereira, A. H. Castro tures (Akademie, Berlin, 1988; Mir, Moscow, 1990).
Neto, and F. Guinea, Rev. Mod. Phys. 84, 1067 (2012).
25. W. A. Harrison, The Electronic Structure and Properties
3. N. Swain and P. Majumdar, arXive: 1610.00695.
of Solids (Freeman, San Francisco, CA, 1980), Vol. 1.
4. M. V. Ulybyshev, P. V. Buividovich, M. I. Katsnelson,
and M. I. Polikarpov, Phys. Rev. Lett. 111, 056801 26. W. A. Harrison, Phys. Rev. B 31, 2121 (1985).
(2013). 27. S. Yu. Davydov and O. V. Posrednik, Bond-Orbital
5. L. Wang, P. Corboz, and M. Troyer, New J. Phys. 16, Method in the Theory of Semiconductors, The Handbook
103008 (2014). (SPbGETU LETI, St. Petersburg, 2007) [in Russian].
twirpx.com/file/1014608/.
6. M. Hohenadler, F. P. Toldin, I. F. Herbut, and
F. F. Assaad, Phys. Rev. B 90, 085146 (2014). 28. S. Yu. Davydov and G. I. Sabirova, Tech. Phys. Lett.
37, 515 (2011).
7. W. Wu and A.-M. S. Tremblay, Phys. Rev. B 89, 205128
(2014). 29. A. Mattausch and O. Pankratov, Phys. Rev. Lett. 99,
076802 (2007).
8. L. Classen, I. F. Herbut, L. Janssen, and M. M. Scherer,
Phys. Rev. B 92, 035429 (2015). 30. J. Borysiuk, J. Soltys, R. Bozek, J. Piechota, S. Kru-
kowski, W. Strupinski, J. M. Baranowski, and R. Step-
9. S. Yu. Davydov, Phys. Solid State 59, 1674 (2017). niewski, Phys. Rev. B 85, 045426 (2012).
10. S. Yu. Davydov, Phys. Solid State 58, 1222 (2016). 31. S. Yu. Davydov, Phys. Solid State 59, 629 (2017).
11. M. Hohenadler and W. von der Linden, Phys. Rev. B 32. S. Yu. Davydov and O. V. Posrednik, Phys. Solid State
71, 184309 (2005). 57, 837 (2015).
12. M. Berciu, Phys. Rev. B 75, 081101R (2007). 33. T. O. Wehling, E. Sasioglu, C. Friedrich, A. I. Lichten-
13. R. P. Hardikar and R. T. Clay, Phys. Rev. B 75, 245103 stein, M. I. Katsnelson, and S. Blügel, Phys. Rev. B
(2007). 106, 236805 (2011).
14. S. Kumar and J. van den Brink, Phys. Rev. B 78, 155123 34. J. C. Johannsen, S. Ulstrup, M. Bianchi, R. Hatch,
(2008). D. Guan, F. Mazzola, L. Hornekær, F. Fromm,
15. Y. Murakami, P. Werner, N. Tsuji, and H. Aoki, Phys. C. Raidel, T. Seyller, and P. Hofmann, J. Phys.: Con-
Rev. B 88, 125126 (2013). dens. Matter 25, 094001 (2013).
16. M. Chakraborty, M. Tezuka, and B. I. Min, Phys. Rev. 35. V. I. Gavrilenko, A. M. Grekhov, D. V. Korbutyak, and
B 89, 035146 (2014). V. G. Litovchenko, Optical Properties of Semiconduc-
17. P. Werner and M. Eckstein, Europhys. Lett. 109, 37002 tors, The Handbook (Nauk. Dumka, Kiev, 1987) [in
(2015). Russian].
18. S. Pradhan and G. V. Pai, Phys. Rev. B 92, 165124
(2015). Translated by E. Smorgonskaya

SEMICONDUCTORS Vol. 52 No. 3 2018

You might also like