Fundamentals of Electronics Book 1 Electronic Devices and Circuit Applications
Fundamentals of Electronics Book 1 Electronic Devices and Circuit Applications
1
V.Lashkariov Institute of Semiconductor Physics, National Academy of Sciences of Ukraine,
pr. Nauky 41, 03028 Kyiv, Ukraine
2
Institute of Physics, National Academy of Sciences of Ukraine,
pr. Nauky 46, 03028 Kyiv, Ukraine
3
Taras Shevchenko Kyiv National University, Radiophysical Faculty
pr. Akademika Hlushkova 4g, 03022 Kyiv, Ukraine
*
corresponding author, e-mail: [email protected]
1
Linear spectrum of graphene causes remarkable symmetry in dependence of its
1, 2, 3.
conductivity on gate voltage: σ (Vg ) = σ ( −Vg ) However, this symmetry occurs in the
highest quality exfoliated graphene with specially prepared surface and interface with substrate.
For less perfect graphene, e.g. for the one produced by CVD and used for applications in
devices, many studies (see e.g. refs. 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18 19, 20, 21, 22, 23) reports about
the hysteresis dependence of graphene conductivity on the gate voltage. The direction of the
hysteresis loop could be different, namely it can be a direct or an inverse ones (anti-hysteresis).
The type of hysteresis and the form of the loop depend on the sweeping rate of the switching
gate Vg, defined as the derivative dVg dt , as well as on the gate voltage range V gsw and on the
Another reason for graphene doping can be polar molecules (e.g. water ones) absorbed on
, 15 - 18, 24, 25, 26, 27
graphene surface . Different substances can cause different type of doping. In
particular absorbed NH3 and CO molecules dope the graphene channel with electrons, and hence
shift of electro-neutrality point to negative voltages range, while H2O and NO2 dope it with holes
and the positive shift of the electro-neutrality point appears.
The doping can be caused by the spontaneous polarization of ferroelectric substrate, at that
the electro-neutrality point shift from V g = 0 can be either right or left depending on the
2
polarization direction. The characteristic hysteresis in the conductivity σ (Vg ) of graphene-on-
Despite the large number of studies on this subject, only a few theoretical models, which
describes hysteresis in GFET and consider the presence of surface dipoles on the free surface of
graphene , as well as interface states at the graphene-substrate , , have been proposed. However,
,
to describe anti-hysteresis in GFET on ferroelectric PZT substrate a quite narrow range of gate
voltages, E<Ec, was considered, and so the contribution of ferroelectric polarization hysteresis
was not taken into account, but only the gate doping given the large dielecric permittivity κ for
PZT.
To resume the review, despite the great amount of efforts focused on the problem, there is
no comprehensive analytical model of hysteretic phenomena in GFETs taking into account all
the abovementioned rival factors. The gap in the knowledge motivated us to evolve a general
theory for the quantitative description of versatile hysteretic phenomena in GFETs on different
substrates. The phenomenological approach presented below considers the existence of both the
external dipoles on graphene free surface and the localized states at the graphene-surface
interface. Our approach takes into account multiple factors, which make a significant
contribution to the conductivity of the graphene channel on substrates of various types, namely:
a) the presence of surface contaminants (dipoles) on the free surface of graphene and their
dynamics depending on the rate of gate voltage switching and relaxation time of dipoles for
GFET systems with ferroelectric and SiO2 substrates; b) the presence of ferroelectric polarization
hysteresis depending on the magnitude of applied electric field considered in the framework of
nonlinear Landau-Khalatnikov relaxation equation; c) the presence of interface states in
graphene-substrate interface.
Theoretical formalism
We study a heterostructure-type system consisting of a conducting graphene channel,
placed on a dielectric or ferroelectric substrate (see Figure 1). The concentration of 2D carriers
in the channel is governed by several factors, which are a time-dependent gate voltage Vg (t ) ,
3
water molecules) on graphene surface, and the charge carriers localized by traps at graphene-
substrate interface. Later we assume that the time dependence Vg (t ) is saw-like Eq. (A.1) with a
period 2π ω and the amplitude A (see Fig.A.1 in the Supplementary Materials 30).
Allowing for all the abovementioned factors, the resistivity of graphene channel can be
presented similarly to :
[ ] σ(V , P1 P , T ) + σ 1(T ) + σ1
ρ Vg , Ps Pf , T ≈ , (1)
g s f intr min
temperature-dependent ferroelectric polarization. In the most common case for graphene under
, 31
ambient conditions the scattering of ionized impurities by substrate dominate , and so the
conductivity of 2D graphene channel can be presented as σ(Vg , Ps Pf , T ) = eµ n (Vg , Ps Pf , T ) ,
where n (V g , Ps , Pf ) is 2D carriers concentration per unit area, caused by the gate mixed
doping, and by the dipoles absorbed by the surface, as well as by ferroelectric dipoles; µ is
carriers mobility that is also a function of temperature T in a general case. The second term in
Eq.(1) corresponds the intrinsic graphene conductivity that should be taken into consideration,
when Fermi level is in the vicinity of electro-neutrality point, σintr (T ) = e µnintr (T ) ,
2(k BT ) 2
nintr (T ) = . The third term in Eq.(1) corresponds the minimal quantum conductivity,
π(hvF ) 2
4
4e 2
σ min ≈ , that should be taken into consideration at low T. Fermi level position in a single-
h
layer graphene is governed by the relation EF = h vF π n(Vg , Ps , Pf ) . Below we consider the
system at room temperature and so omit the T-dependence for the sake of brevity.
dVg (t )
Later we’ll consider at first the forward sweep in Eq.(A.1, Appendix A), > 0 and
dt
start from the case without any dipoles, Pf = 0, Ps = 0 . However, we assume, that the localized
states with the energy ET and concentration nT are present at the graphene-substrate interface. For
a voltage range V g (t ) < VT 1 , where VT1 corresponds to the situation when the occupation of the
interface states with the electrons from graphene channel starts, E F (VT 1 ) = ET 1 , the 2D
concentration of electrons in graphene channel is governed by a simple capacitor formula , :
κVg (t )
n(Vg , t ) = , (2)
4πed
where κ is the dielectric permittivity of substrate and d is the substrate thickness. The gate
voltage VT1 leads to the start of the interface states occupation with electrons from the graphene
channel ( E F (Vg (t )) = ET 1 ). The voltage VT2 corresponds to the situation, when all localized
interface states are already occupied by electrons from the graphene channel. The voltages are
determined by expressions , :
4πed ET21 4πed ET22 4πednT
VT 1 = , VT 2 = + . (3)
κ πh 2 v F2 κ πh v F
2 2
κ
In the voltage range VT 1 ≤ V g < VT 2 , for which the occupation of interface states occur, the
concentration in graphene channel remains constant, because additional free carriers, injected
from contacts into the graphene channel, are immediately captured by the interface states , and
ET21
so n = . In the voltage range VT 2 ≤ Vg where all the interface states are already occupied
πh 2 v F2
by electrons, the free electrons' concentration in the channel is governed by the evident relation
κ V g (t )
n(V g ) = − nT .
4πed
Let us take into consideration the polarization of dipoles both in the ferroelectric substrate
, ,
and at the graphene surface. If Ps(t) ≠ 0, Pf(t, T) ≠ 0, E F ≤ ET 1 , similarly to we get:
κVg (t ) Ps (t ) + Pf (t , T )
n (Vg , Ps , Pf ) = + (4)
4πe d e
5
Below we describe the temperature dependent spontaneous ferroelectric polarization by
commonly used expression 32:
Pf (t , T ) = Pf (T ) tanh (s f (V g (t ) − Vc )) , (5)
where Vc = Ec d is a coercive voltage equal to the product of coercive field and substrate
⎡ ⎛ t (Vs ) − t ) ⎞⎤
Pf (t , T ) = Pf (T ) tanh (s f (V g (t ) + Vc )) and Ps (t ) = Ps ⎢1 − exp⎜ − ⎟⎥ , where t (Vs ) is the
⎣ ⎝ τ ⎠⎦
moment of time, corresponding to the complete suppression of Ps by the critical gate voltage, τ
is the dipoles' relaxation time that can be rather long of several seconds order, because the
absorbed dipoles are obviously non-Langevin ones due to the chemical bonding to graphene
surface . Hereinafter we assume that the polarization of the absorbed dipoles starts its renovation
immediately after the switching to the backward sweep.
We also assume, that these dipoles shift the electro-neutrality point into the positive range of Vg ,
6
as the water molecules do. This means the conductivity of graphene channel is determined by
holes that at zero gate voltage.
For this case the carriers concentration in graphene channel is determined both by gate
doping and by the absorbed dipoles polarization, which in turn depends on the external electric
field, caused by the gate voltage. The carriers concentration is as large as
nPs = Ps (t ) e ≈ 6.23 × 1013 cm −2 for the polarization value Ps = 0.1 C/m2. Therefore it exceeds in
two orders of magnitude the maximal concentration nGD = κ V g 4 π e d ≈ 4.42 × 1011 cm −2 caused
by the gate doping for the range −8V <Vg<+8V. For the estimation we used the most common
parameters of SiO2 substrate κ = 3.9 and d = 300 nm. Therefore the conductivity of the
graphene channel would be governed by time-dependent polarization of the dipoles.
Figure 2 shows the gate voltage dependence of carrier concentration in the graphene
channel on SiO2 substrate calculated from Eqs. (4), (6), (A.1) and (A.18) from Suppl. Mat. for
different times of surface dipoles relaxation τ and the same switching time. In the initial moment
of time (at Vg = 0) graphene’s channel conductivity is determined by holes, as it was noted
before. However, the electric field, caused by the gate potential, destroys the dipoles polarization
at some critical value of Vg. Note, that the dependences in Figure 2 are presented after the first
complete destruction of the surface dipoles polarization; solid lines correspond to the second
forward sweep, i.e. polarization disappearance after the first cycle of its renewal. The
polarization will start to renew at the backward sweep. The value of polarization that recovers
within a switching period is different for different relaxation times τ., but it is clear from the
figure, that the polarization renews almost completely within a switching period for a short
relaxation time. Therefore the concentration of 2D carriers in graphene is determined by the
polarization, and a pronounced hysteresis loop, although asymmetrical one, is characteristic for
n(Vg ) dependence.
In contrast, the polarization cannot be renewed after being destroyed for the first time for
long enough relaxation times, since the 2D carriers concentration is determined by gate doping
Eq.(2), and it is, in orders of magnitude smaller according to the previous estimation. Mention,
that a direction of such hysteresis loop would be opposite to the so-called direct one, created by
ferroelectric dipoles (“anti-hysteresis”).
7
Figure 2. The dependence of carrier concentration in the graphene channel on SiO2 substrate vs.
the gate voltage calculated for different times τ of surface dipoles relaxation. Forward and
backward sweeping dependences are presented by solid and dotted curves respectively. The
curves, calculated for different dipoles' relaxation times τ=1, 5, 10 and 50 s, are shown by
different colours. Parameters, used in calculations are κ=3.9, d = 300 nm, Vs = 0.3 V,
dVg dt = +5 V/s for the forward sweep and dVg dt = −5 V/s for the backward one, Vgmax = ±6 V,
Figure 3 presents the dependence of 2D concentration in graphene channel vs. the gate
voltage as a function of the gate voltage switching range. The figure 3 shows that the hysteresis
region increases under the gate voltage increase. Considering the direct sweeping (at
dVg dt > 0 ), it is seen that the disappearance of surface dipoles' polarization occur for all ranges
of gate voltages at fixed electric field (see solid curves and Eqs. (4), (6)). After switching the
gate voltage in the reverse direction, the dipoles' polarization on the graphene surface
immediately starts to recover (see dotted lines and Eq. (A.18)). Therefore, it becomes clear that
the value of the polarization that has time to recover will be different for different gate voltage
ranges at the same sweeping rate dVg dt , because the time required to reach different voltage
values (-2V, -4V, -6V, -8V) will be different and corresponding polarization value that will have
time to recover will be bigger for a higher gate voltage Vg.
8
Figure 3. The dependence of carrier concentration in the graphene channel on SiO2 substrate vs.
the gate voltage calculated for different switching ranges for the gate voltage. Forward and
backward sweeping dependences are presented by solid and dotted curves respectively. The
curves, calculated for different sweeping range Vgmax = ± 2V, ± 4V, ± 6V and ± 8V, are
shown by different colours. Parameters used in calculations are κ=3.9, d = 300 nm, Vs = 0.3 V, τ
= 5 s, dVg dt = ± 1 V/s; Ps = 0.1 C/m2 and ss = 1.
B. Perfect ferroelectric substrate without interface states between the channel and the
substrate. As the next step let us consider the graphene channel on an ideally perfect Pb(ZrxTi1-
x)O3 (PZT) substrate without interface states between the channel and the substrate. PZT is
, ,
widely used as substrate for GFETs , because its permittivity varies in a broad range
depending on the chemical composition x, and can be very high near the morphotropic phase
boundary at х = 0.52.
It is crucial for our consideration, that PZT response is not a hysteretic one for the electric
fields smaller than some coercive value, Ec, for which it can be presented by a linear dependence
P (V g ) (see works ,
and refs. therein). Thus below we will operate within a relatively narrow
gate voltage switching range from – 8V to + 8V for which the linear response approximation
P (V g ) is valid. Therefore the further analysis should be similar to the one considered above, but
the ferroelectric substrate permittivity value is much higher than unity. For κ = 400 the carrier
concentration n(Vg ) , caused by the gate doping, would be of the same order as the n(Vg ) caused
by the “doping by absorbed surface dipoles” with Ps = 0.1 C/m2. The maximal value of n(Vg ) for
the switching range Vgsw = ±8 V can be as large as nGD = κ V g 4 π e d ≈ 5.89 × 1013 cm −2 , while
the doping by the surface dipoles can be nPs ≈ 6.23 × 1013 cm −2 . However, the gate doping creates
electron and hole conductivity that depends on the gate voltage polarity, while the doping by
9
surface dipoles for the case of the absorbed water molecules creates the hole conductivity only.
Therefore the total concentration of 2D carriers would be determined by combination of the
“gate” and “surface dipoles” doping according to Eq.(A.12), (A.14) and (A.18) in Ref..
Figure 4 presents the carrier concentration in graphene calculated for different dipoles'
relaxation times τ. Similarly to the case of SiO2 substrate, the graphene channel on PZT substrate
possess the hole conductivity at Vg = 0. With a shift of the gate voltage into a positive range the
holes concentration decreases and finally reaches the electro-neutrality point (VNP1= 1.76 V for τ
= 1 s). In this point the graphene valence band is completely occupied by electrons, and the
conduction band is empty. With a further increase of the gate voltage the graphene channel
conductivity switches to an electron one, and at the switching point Vg = 6V the electrons
concentration is n = 4.5 × 1013 cm −2 .
Figure 4. The dependence of carrier concentration in the graphene channel on PZT substrate vs.
the gate voltage calculated for different relaxation times of surface dipoles. Forward and
backward sweeping dependences are presented by solid and dotted curves respectively. The
curves, calculated for different dipoles' relaxation times τ=1, 5, 10 and 50 s, are shown by
different colours. Parameters used in calculations are Vs = 0.3 V, dVg dt = ± 5 V/s, Vgmax = ± 6
For the backward sweep the electro-neutrality point would occur at higher voltages for a
short relaxation time of surface dipoles (VNP2 = 3.5 V for τ = 1 s) than for the forward one (VNP1=
1.76 V for τ = 1 s). With a further change of the gate voltage up to Vg = – 6V, the conductivity of
the graphene channel will be determined by holes with concentration n ≈ 1014 cm −2 . With the
next forward and backward sweeps the hysteresis loop would be reproducible under the same
10
other conditions. The direction of the hysteresis loop, presented in Figs.2-4, is opposite to one
another, determined by the ferroelectric dipoles polarization, namely the electro-neutrality point
for the backward sweep NP2 shifts to the right from NP1. Under long relaxation times τ ≥ 50 s
the surface polarization cannot restore and corresponding concentration dependence on the gate
voltage is described by a linear dependence Eq.(2).
Figure 5 presents the dependence of carriers' concentration in the graphene channel on
PZT substrate vs. the gate voltage calculated for its different amplitude Vgmax . As anticipated
these dependences shows anti-hysteresis behavior and they are qualitatively similar to those for
graphene on SiO2 substrate.
Figure 5. The dependence of carrier concentration in the graphene channel on PZT substrate vs.
the gate voltage for different amplitudes of gate voltage switching. Forward and backward
sweeping dependences are presented by solid and dotted curves respectively. Parameters used in
calculations are d = 300 nm, κ = 400, Vs = 0.3 V, τ = 5 s, dVg dt = ± 1 V/s, Ps = 0.1 C/m2, ss =
C. Ideal ferroelectric substrate without surface dipoles. Let us consider the GFET on an
ideal ferroelectric substrate without interface states (as it was done in 34) and without any surface
dipoles, which can be removed by e.g. annealing . In this case Ps = 0, Pf ≠ 0, nT = 0 , and so we
can take into account the ferroelectric polarization response on the gate voltage only. This
35
response can be described using Landau-Khalatnikov (LK) relaxation equation that is also
called time-dependent Ginzburg-Landau equation 36. It reads
11
dP ( t )
Γ = α (T )P (t ) + β P 3 (t ) + γ P 5 (t ) − E (ω, t ) (7)
dt
where the Landau potential expansion coefficients α = 4.24 × 10 5 (T − 691) Jm/C2,
β = 4 × 0.3614 × 10 8 C-4·m5J and γ = 6 × 1.859 × 108 Jm9C-6. All further calculations are performed
at room temperature (T=293 К) when the LK relaxation time τ LK = Γ α is very small and has a
phonon time order ~(10-11 – 10-13) s. Note that LK equation can describe the ferroelectric
polarization relaxation dynamics in the system "graphene-on-ferroelectric substrate" 37.
Numerical computations according to Eq.(7) enable to obtain the coercive field E c (E )
dependence on applied electric field E = Vg d created by the gate voltage (Figure 6).
Figure 6. Coercive voltage Vc as a function of gate voltage Vgate calculated for ferroelectric
substrate thickness d = 300 nm according to Eq.(7).
switching with a change of the gate voltage from the maximal negative value − Vgmax to the
positive one + Vgmax . Note, that a distance between electro-neutrality points increases with
increasing of the voltage switching range. This hysteresis is caused by the two different stable
, ,
polarizations of ferroelectric crystal, it can be used for advanced memory cells (see reviews ).
However, the pronounced hysteresis can be observed up to the switching times comparative with
the times of ferroelectric domains reversal. This imposes rather strong restrictions on the
12
operating frequency (104 - 106 Hz) for bulk ferroelectric crystals. However, these times can be
several orders of magnitude smaller (about 10–9s) for thin ferroelectric films of µm thickness and
thinner . In our case the nonlinear ferroelectric response will take place at all the gate voltages
switching rates within the considered range (0.1 − 50) V/s. This means that the ferroelectric
dipoles re-polarization will always follow the change of external field, as well as the hysteresis
loop width is constant for a given switching range and does not depend on the sweeping rate.
Such a situation was observed experimentally for graphene on PTO/STO substrate with SRO
gate .
Figure 7. The dependence of carrier concentration in the graphene channel on ideal ferroelectric
substrate vs. gate voltage calculated for different amplitudes of the gate voltage switching Vgmax .
Forward and backward sweeping dependences are presented by red and black curves
respectively. Parameters used in calculations are κ =400, Pf = 0.1 C/m2, d = 300 nm, Vgmax =
D. Gate doping together with doping by ferroelectric substrate and surface dipoles. Now
we can modify the previous case "C" by taking into account surface dipoles as well. Therefore
we consider the case Ps ≠ 0, Pf ≠ 0, nT = 0 , so the two rival mechanisms of doping coexist, the
first one originates from ferroelectric dipoles and the second one steams from absorbed surface
dipoles.
Let us chose the switching range from – 6V to + 6V, for which the ferroelectric response
have a form of hysteresis loop with a width Vc = 0.4 V, and take the values of ferroelectric and
surface dipoles polarization as Ps = 0.1 C/m2 and Pf = 0.1 C/m2 correspondingly. Since the value
of ferroelectric permittivity κ = 400 is high enough, the carriers concentration nGD , caused by
the gate doping for Vg = ±6 V , is of the same order of magnitude than the concentration induced
13
by ferroelectric and surface dipoles nPs . In particular, nGD = κVg 4 π e d ≈ 4.42 × 1013 cm−2 and
Figure 8. The dependence of carriers concentration in the graphene channel with surface
absorbed dipoles on real ferroelectric substrate on gate voltage for different rates of gate voltage
sweeps. Forward sweep curves are red, the backward ones are black. Parameters, used in the
calculations are d = 300 nm, κ =400, Vs = 0.2 V, Vc = 0.4 V, τ = 1 s, Vgmax = ± 6 V, Ps = 0.1
C/m2, Pf = 0.1 C/m2, sf = 1, ss = 1, dVg dt = ± 0.5 V/s (a), dVg dt = ± 50 V/s (b).
As one can see from Figure 8, the existence of two rival channels of graphene doping
modifies the form of hysteresis loops, presented in previous figures. As it could be expected, the
effect of the absorbed surface molecules is essential as slow gate voltage sweeps, and so called
anti-hysteresis occurs (Fig.8a). With the increase of dVg dt the anti-hysteresis vanishes and
finally it transfers into the ferroelectric “direct hysteresis” with the increase of the distance
between the electro-neutrality points, up to its final saturation (Fig. 8b).
38
Figure 9 presents experimental and theoretical values of electro-neutrality point
positions as the function of gate voltage sweeping rate for GFET on PZT ferroelectric substrate.
In the experiment a single-layer graphene was fabricated by CVD method on cooper film, and
then deposited onto the clean surface of 140 nm PbZr0.2Ti0.8O3 /60 nm SrRuO3/ SrTiO3 (001)
heterostructure. Next the palladium contacts to graphene were launched. As one can see from
Figure 9, our theory is in a good agreement with experimental data . Small differences between
the theoretical and experimental dependences can be caused by the interface states between
graphene and ferroelectric, as well as by chemical doping of graphene during its fabrication.
14
Figure 9. Experimental and theoretical values of electro-neutrality point positions as the
function of gate voltage sweeping rate for GFET on PZT ferroelectric substrate. Parameters,
taken from are: d=200 nm, Vc = 2V, Pf = 0,23 C/m2, Vgmax ± 5V, for each of the five points
dVg dt = 0.0084 V/s, 0.017 V/s, 0.084 V/s, 1 V/s, 2.87 V/s, τ = 20 s. Fitting parameters are Vs =
E. Gate doping together with doping by ferroelectric substrate and trapping by the
interface states. Finally we can analyze the more general case, Ps = 0, Pf ≠ 0, nT ≠ 0 , assuming
that the concentration of the interface states is high enough to trap all the electrons injected into
the channel due to the gate doping and doping by ferroelectric entire considered Vg-range (the
validity of this assumption was discussed in , ). We also assume that there is no saturation of
ferroelectric permittivity hysteresis for this range.
Fermi energy in a single layer graphene is governed by expression E F = h v F π n . Let us
take into consideration the possibility of the carriers trapping from graphene channel by localized
states connected with the graphene-surface interface. These states are described by the definite
density of states ρT (E). We assume that ρT (E) is non-zero in some energy interval ET1 <E <ET2
only, and is equal to zero out of the range.
Figure 10 demonstrates hysteresis of free carrier concentration n(Vg) in the graphene
channel on ferroelectric substrate vs. the gate voltage. The dependences were calculated
according to Eqs. (A.12), (A.15a)-(A.16b) in Ref.. using the dependence Pf (Vg ) given by Eqs.
15
The form of n(Vg) would be determined in this case by competition of the two rival mechanisms
(ferroelectric doping and charge trapping). For the considered case a graphene channel has the
holes conductivity at zero gate voltage due to ferroelectric hysteresis. The increase of Vg leads to
the decrease of n(Vg) until the electro-neutrality point B would be reached. In this point the
graphene valence band is completely occupied by electrons, and the conduction band is empty.
With a further increase of the gate voltage the graphene channel conductivity switches to an
electron one (B→C) and the concentration n(Vg) would increase until the moment, when the
Fermi level would reach the interface states energy (VT1, point С), and the electrons trapping by
these states starts. In the range VT 1 ≤ V g < VT 2 (C→D) where the trapping occurs, the
concentration of the free carriers in the graphene channel would be constant, because all the
additional carriers, injected into the channel trough the contacts due to the gate and ferroelectric
doping, are almost immediately captured by the interface states. After all, i.e. at VT 2 ≤ Vg (t ) , the
interface states would be occupied by electrons and so the dependence n(Vg) would be governed
by Eq.(A.16b). Since the lifetime of carriers on the interface states is much longer than the
switching time, the backward sweep (D→E→F→G→A) can include both the direct hysteresis
(Fig. 10a) and the inverse one (“anti-hysteresis”, Fig.10b). Everything is determined by the
number of carriers, trapped by interface states, which negative charge screens the electric field in
the substrate, that in turn leads to the right-hand-side shift of electroneutrality point along the
gate voltage scale, which appears during the backward switching , , the energy of interface states
and the ferroelectric polarization. On the backward sweep graphene would again change its type
of conductivity to a hole one (E→F region); in the point F the depletion of the interface states
will start, and it would be accomplished in the point G. Finally the hysteresis loop would be
closed. For numerical simulations we used the following parameters of the localized states for
the electrons trapping and depletion, ET1=0,55 eV, ∆E = ET2 − ET1= 50 µeV, ET1(hole)=0,83 eV
and ∆E(hole)=ET2(hole) − ET1(hole)= 50 µeV. The concentration of electrons, trapped by interface
states, is n T = 6.75 × 1013 cm −2 (a) and n T = 9.16 × 1013 cm −2 (b).
16
Figure 10. Hysteresis dependence of free carrier concentration in graphene channel on
ferroelectric substrate vs. the gate voltage. The direction of loop is
A→B→C→D→E→F→G→A. The parameters d = 300 nm, κ=400, Pf = 0,1 C/m2, Vgmax ± =4V
(a) and Vgmax ± =6V (b). Forward sweeping dependences are shown by red curves, the backward
Conclusion
Using a phenomenological approach, we developed a general theory for the analytical
description of versatile hysteretic phenomena in GFETs on different substrates. Evolved
formalism considers different origin of the existence of the external dipoles on graphene free
surface and the localized states at the graphene-surface interface.
We demonstrated that the dipole molecules absorbed by the graphene surface (e.g. polar
water molecules) can cause hysteretic form of the graphene conductivity dependence on a gate
voltage n(Vg) in GFET on different substrates. It was demonstrated, that the increase of the gate
voltage sweeping rate dV g dt leads to the complete vanishing of hysteresis both for GFET on
SiO2 substrate and on ferroelectric substrate for the gate voltages less than the coercive one,
Vg<Vc. These results are in qualitative and quantitative agreement with the available
experimental data . Also we have shown that the increase of gate voltage rate causes the
17
transition from anti-hysteresis to ferroelectric hysteresis for PZT substrate, that correlates well
with the experiment .
Notice that our results have been obtained within a number of essential approximations and
simplifications, such as the approximation for spontaneous relaxation of the absorbed dipoles
polarization with a single relaxation time τ. More rigorous model should be based on deeper
understanding of the nature of dipoles bonding to graphene surface and account for a possible
relaxation time spectrum. Also we treated the process of the carriers trapping by surface states as
a very swift process independent on the frequency of gate voltage switching. However, it was
demonstrated for GFET on PbTiO3/SrTiO3 substrate in ref., that the process of carriers trapping
becomes less effective for the gate voltage switching frequency higher than 1 KHz that
corresponds to the sweeping rate 104 V/s. The understanding of the effect is possible on the base
of the clear vision of the physical mechanism of trapping and the configuration of its potential.
Since GFETs on ferroelectric substrate are very promising candidates for the non-volatile
, ,
ultra-fast memory of new generation (see e.g. ), the theoretical description of the devices
operation requires reliable knowledge about the ferroelectric response at low, intermediate and
high frequencies, as well as about the processes of carrier trapping at a given frequency range.
This knowledge is crucial, because the competition of the abovementioned factors would
determine the FRAM operation characteristics. Obtained results are valid for the description of
hysteretic phenomena in various realistic GFETs operating at low and intermediate frequency
range, may be not directly applicable for the description of ferroelectric response at ultra-high
frequencies, priory because the high-frequency response of ferroelectric dipoles requires further
experimental studies.
18
References
1
Novoselov, K.S.; Geim, A.K.; Morozov, S.V.; Jiang, D.; Zhang, Y.; Dubonos, S.V.; Grigorieva, I.V.;
Firsov, A.A. Electric Field Effect in Atomically Thin Carbon Films. Science 2004, 306, 666 – 669.
2
Geim, A. K. Graphene: Status and Prospects. Science 2009, 324, 1530 – 1534.
3
Das Sarma, S.; Adam, S.; Hwang, E. H.; Rossi, E. Electronic transport in two-dimensional graphene.
Rev. Mod. Phys. 2011, 83, 407 – 470.
4
Raghavan, S.; Stolichnov, I.; Setter, N.; Heron, J.-S.; Tosun, M.; Kis, A. Long-term retention in organic
ferroelectric-graphene memories. Appl. Phys. Lett. 2012, 100, 023507.
5
Zhang, X.-W.; Xie, D.; Xu, J.-L.; Zhang, C.; Sun, Y.-L.; Zhao, Y.-F.; Li, X.; Li, X.-M.; Zhu, H.-W.;
Chen, H.-M.; Chang, T.-C. Temperature-dependent electrical transport properties in
graphene/Pb(Zr0.4Ti0.6)O3 field effect transistors. Carbon 2015, 93, 384 –392.
6
Kalon, G.; Shin, Y. J.; Truong, V. G.; Kalitsov, A.; Yang, H. The role of charge traps in inducing
hysteresis: Capacitance–voltage measurements on top gated bilayer graphene. Appl. Phys. Lett. 2010, 99,
083109.
7
Tan, J. Y.; Avsar, A.; Balakrishnan, J.; Koon, G. K. W.; Taychatanapat, T.; O'Farrell, E. C. T.;
Watanabe, K.; Taniguchi, T.; Eda, G.; Castro Neto, A. H.; Özyilmaz, B. Electronic transport in graphene-
based heterostructures. Applied Physics Letters 2014, 104, 183504.
8
Rajapitamahuni, A.; Hoffman, J.; Ahn, C. H.; Hong, X. Examining Graphene Field Effect Sensors for
Ferroelectric Thin Film Studies, Nano Letters 2013, 13, 4374-4379.
9
Sangwan, V. K.; Jariwala, D.; Everaerts, K.; McMorrow, J. J.; He, J.; Grayson, M.; Lauhon, L. J.;
Marks, T. J.; Hersam, M. C. Wafer-scale solution-derived molecular gate dielectrics for low-voltage
graphene electronics. Applied Physics Letters 2014, 104, 083503.
10
Sachs, R.; Lin, Z.; Shi, J. Ferroelectric-like SrTiO3 surface dipoles probed by graphene. Scientific
Reports 2014, 4, 3657.
11
Wang, H.; Wu, Y.; Cong, C.; Shang, J.; Yu, T. Hysteresis of Electronic Transport in Graphene
Transistors. ACS Nano 2010, 4, 7221–7228.
19
12
Joshi, P.; Romero, H.E.; Neal, A. T.; Toutam, V. K.; Tadigadapa, S. A. Intrinsic doping and gate
hysteresis in graphene field effect devices fabricated on SiO2 substrates. J. Phys.: Condens. Matter 2010,
22, 334214.
13
Liao, Z.-M.; Han, B.-H.; Zhou, Y.-B.; Yu, D.-P. Hysteresis reversion in graphene field-effect
transistors. J. Chem. Phys. 2010, 133, 044703.
14
Hong, X.; Hoffman, J.; Posadas, A.; Zou, K.; Ahn, C. H.; Zhu, J. Unusual resistance hysteresis in n -
layer graphene field effect transistors fabricated on ferroelectric Pb(Zr0.2 Ti0.8) O3. Applied Physics Letters
2010, 97, 033114.
15
Lafkioti, M.; Krauss, B.; Lohmann, T.; Zschieschang, U.; Klauk, H.; Klitzing, K.; Smet, J. H. Graphene
on a hydrophobic substrate: doping reduction and hysteresis suppression under ambient conditions. Nano
Lett., 2010, 10, pp 1149–1153.
16
Sabri, S. S.; Lévesque, P. L.; Aguirre, C. M.; Guillemette, J.; Martel, R.; Szkopek, T. Graphene field
effect transistors with parylene gate dielectric. Appl. Phys. Lett. 2009, 95, 242104.
17
Levesque, P. L.; Sabri, S. S.; Aguirre,C. M.; Guillemette, J.; Siaj, M.; Desjardins, P.; Szkopek, T.;
Martel, R. Probing charge transfer at surfaces using graphene transistors. Nano Lett. 2011, 11, 132–137.
18
Veligura, A.; Zomer, P. J.; Vera-Marun, I. J.; Józsa, C.; Gordiichuk, P. I.; van Wees, B. J. Relating
hysteresis and electrochemistry in graphene field effect transistors. J. Appl. Phys. 2011, 110, 113708.
19
Zheng, Y.; Ni, G.-X.; Bae, S.; Cong, C.-X.; Kahya, O.; Toh, C.-T.; Kim, H. R.; Im, D.; Yu, T.; Ahn, J.
H.; Hong, B. H.; Ozyilmaz, B. Wafer-scale graphene/ferroelectric hybrid devices for low-voltage
electronics. EPL 2011, 93, 17002.
20
Song, E. B.; Lian, B.; Kim, S. M.; Lee, S.; Chung, T.-K.; Wang, M.; Zeng, C.; Xu, G.; Wong, K.;
Zhou, Y.; Rasool, H. I.; Seo, D. H.; Chung, H.-J.; Heo, J.; Seo, S.; Wang, K. L. Robust bi-stable memory
operation in single-layer graphene ferroelectric memory. Applied Physics Letters. 2011, 99, 042109.
21
Strikha, M. V. Non-volatile memory and IR radiation modulators based upon graphene-on-ferroelectric
substrate. A review, Ukr. J. Phys. Opt. 2012, 13, S5-S27.
22
Strikha, M. V. Non Volatile Memory of New Generation and Ultrafast IR Modulators Based on
Graphene on Ferroelectric Substrate. In Functional Nanomaterials and Devices for Electronics, Sensors
20
and Energy Harvesting; Nazarov, A., Balestra, F., Kilchytska, V., Flandre, D., Eds.; Springer
International Publishing: Switzerland, 2014; pp 163–178.
23
Strikha, M. V. Hysteresis in the Resistivity of Graphene Channel. In: Chemical Functionalization of
Carbon Nanomaterials: Chemistry and Application; Thakur, V. K., Thakur, M. K., Eds.; Taylor and
Francis: New York, 2015, pp 939-948.
24
Schedin, F.; Geim, A. K.; Morozov, S. V.; Hill, E. W.; Blake, P. Katsnelson, M. I.; Novoselov, K. S.
Detection of individual gas molecules adsorbed on graphene. Nature Materials 2007, 6, 652-655.
25
Kong, L.; Enders, A.; Rahman, T. S.; Dowben, P. A. Molecular adsorption on graphene. J. Phys.:
Condens. Matter 2014, 26, 443001.
26
Кurchak, A. I.; Morozovska, A.N.; Strikha, М.V. Rival mechanisms of hysteresis in the resistivity of
graphene channel. Ukr. J. Phys. 2013, 58, 472-479.
27
Shin, D.-W.; Lee, H. M.; Yu, S. M.; Lim, K.-S.; Jung, J. H.; Kim, M.-K.; Kim, S.-W.; Han, J.-H.;
Ruoff, R. S.; Yoo, J.-B. A facile route to recover intrinsic graphene over large scale. ACS Nano 2012, 6,
7781-7788.
28
Strikha, M.V. Mechanism of the antihysteresis behavior of the resistivity of graphene on a Pb(ZrxTi1–
x)O3 ferroelectric substrate. JETP Letters 2012, 95, 198-200.
29
Kurchak, A.I.; Strikha, M.V. Antihysteresis of the electrical resistivity of graphene on a ferroelectric
Pb(ZrxTi1– x)O3 substrate. JETP 2013,143, 129–135.
30
Supplementary Materials [URL will be provided by Publisher]
31
Krugliak, Y. O.; Strihka, M.V. Generalized Landauer–Datta–Lundstrom Model in Application to
Transport Phenomena in Graphene. Ukr. J. Phys. Reviews 2015, 10, 3-32.
32
Tagantsev, A.K.; Cross, L. E.; Fousek, J. Domains in ferroic crystals and thin films. Springer: New
York, 2010, pp 1-821.
33
Zheng, Y.; Ni, G.-X.; Toh, C.-T.; Tan, C.-Y.; Yao, K.; Ozyilmaz, B. Graphene field-effect transistors
with ferroelectric gating. Phys. Rev. Lett. 2010, 105, 166602.
21
34
Yusuf, M.H.; Nielsen, B.; Dawber, M.; Du, X. Extrinsic and Intrinsic Charge Trapping at the
Graphene/Ferroelectric Interface. Nano Letters 2014, 14, 5437-5444.
35
Landau, L.D.; Khalatnikov, I. M. On the anomalous absorption of sound near a second order phase
transition point. In Dokl. Akad. Nauk SSSR; 1954, pp 469-472.
36
Kalinin, S.V.; Morozovska, A.N.; Chen, L. Q.; Rodriguez, B. J. Local polarization dynamics in
ferroelectric materials. Rep. Prog. Phys. 2010, 73, 056502-1-67.
37
Morozovska, A. N.; Pusenkova, A. S.; Varenyk, O. V.; Kalinin, S. V.; Eliseev, E. A.; Strikha, M. V.
Finite size effects of hysteretic dynamics in multi-layer graphene on ferroelectric. Physical Review B
2015, 91, 235312.
38
Baeumer, C.; Rogers, S. P.; Xu, R.; Martin, L. W.; Shim, M. Tunable carrier type and density in
graphene/PbZr0.2Ti0.8O3 hybrid structures through ferroelectric switching. Nano Letters. 2013, 13, 1693–
1698.
39
Park, N.; Kang, H.; Park, J.; Lee, Y.; Yun, Y.; Lee, J.-H.; Lee, S.-G.; Lee, Y. H.; Suh, D. Ferroelectric
single-crystal gated graphene/hexagonal-BN/ferroelectric field effect transistor. ACS Nano, 2015, 9,
10729–10736.
22
Supplementary Materials to
"Hysteretic phenomena in GFET: general theory and experiment"
1
V.Lashkariov Institute of Semiconductor Physics, National Academy of Sciences of Ukraine,
pr. Nauky 41, 03028 Kyiv, Ukraine
2
Institute of Physics, National Academy of Sciences of Ukraine,
pr. Nauky 46, 03028 Kyiv, Ukraine
3
Taras Shevchenko Kyiv National University, Radiophysical Faculty
pr. Akademika Hlushkova 4g, 03022 Kyiv, Ukraine
series:
Figure A1. Schematic saw-like dependence of gate voltage on time (all in arbitrary units).
*
corresponding author, e-mail: [email protected]
1
The resistivity of graphene channel can be presented similarly to 1 as:
ρ (Vg (t ), Ps (t ), Pf (t , T ), T ) ≈
1 1 1
+ + , (A.2)
σ (Vg (t ), Ps (t ), Pf (t , T )) σ int r (T ) σ min
where n (Vg (t ), Ps (t ), Pf (t , T )) is the 2D concentration of carriers per unit area, caused my mixed
doping by gate, surface absorbed dipoles and ferroelectric dipoles, µ is the carriers mobility, which
also are the functions of temperature T.
The second term in Eq.(A.2) corresponds the intrinsic graphene conductivity, which should be
taken into consideration if the Fermi level is close to the electro-neutrality point:
σintr (T ) = e nintr (T )µ (A.4)
2(kT ) 2
nintr (T ) = . (A.5)
π(hvF ) 2
The third term in Eq.(A.2) corresponds the minimal quantum conductivity, which is to be taken
into consideration at low T.
4e 2
σ min ≈ . (A.6)
h
Fermi energy in a single-layer graphene is given by relation 2:
E F = h v F (π n(Vg (t ), Ps (t ), Pf (t , T )))
12
(A.7)
dVg (t )
At first let's analyze the forward sweep in Eq.(A.1), > 0 , for the case without any
dt
dipoles, Pf = 0, Ps = 0 . Also we assume, that the localized states with energy ET and concentration nT
are present at the graphene-substrate interface. For voltage range Vg (t ) < VT 1 , where VT1 corresponds
to the condition E F (VT 1 ) = ET 1 (the occupation of the interface states with the electrons from
graphene channel starts), the 2D concentration of electrons in graphene channel is governed by a
simple capacitor formula 3, 4:
κVg (t )
n(Vg (t )) = (A.8)
4πe d
where κ is dielectric permittivity and d is the thickness of substrate The gate voltage VT1 leads to the
start of the interface states occupation with electrons from the graphene channel ( EF (Vg (t )) = ET 1 ).
2
The voltage VT2 corresponds to the situation, when all localized interface states are already occupied
by electrons from the graphene channel. The voltages 1, :
4πed ET21 4πed ET22 4πednT
VT 1 = , VT2 = + (A.9)
κ πh v F
2 2
κ πh v F
2 2
κ
In the voltage range VT 1 ≤ Vg < VT 2 , where the occupation of interface states occur, the
concentration in graphene channel remains constant, because additional free carriers, injected from
contacts into the graphene channel, are immediately captured by interface states4:
ET21
n= 2 2 (A.10)
πh v F
For the range VT 2 ≤ Vg , when all the interface states are already occupied by electrons, the free
where Vc = Ec d is coercive voltage, equal to the product of coercive field and ferroelectric thickness
d, sf is a fitting parameter, correspondent to the ‘sharpness’ of ferroelectric switching.
The spontaneous polarization of the dipoles absorbed by graphene surface can be presented
as:
1 − tanh(s s (Vg (t ) − Vs ))
Ps (t ) = Ps (A.14)
2
where Vs is some critical voltage (the increase of gate voltage to some critical value, dependent on
system geometry, finally ruins this polarization ), ss is a fitting parameter, correspondent to the
‘sharpness’ of dipoles switching.
Taking into account these two types of polarization, Eq. (A.9-11) can be rewritten as:
4πed ET21 ( Ps (t ) + Pf (t , T ))
VT 1 = [ 2 2 − ] , (A.15a)
κ πh v F e
3
4πed ET22 ( Ps (t ) + Pf (t , T ))
VT 2 = [ 2 2 + nT − ], (A.15b)
κ πh v F e
ET21 Ps (t ) + Pf (t , T )
n= 2 2 − , (A.16a)
πh v F e
κV g Ps (t ) + Pf (t , T )
n(V g ) = + − nT (A.16b)
4πed e
Note, that Eqs.(A.15)-(A.16) are valid for the case, when a life time of carriers, trapped by
interface states, is much greater than the switching time. The validity of the approximation for
graphene on PZT substrate was demonstrated experimentally in 5.
Eqs. (A.15)-(A.16) are valid both for the forward sweep ( dVg (t ) dt > 0 ), and for the backward
one ( dVg (t ) dt > 0 ). However, for the backward sweep Ps and Pf should be obviously presented as:
⎡ ⎛ t ( Vs ) − t ⎞ ⎤
Ps ( t ) = Ps ⎢1 − exp⎜ − ⎟⎥ (A.18)
⎣ ⎝ τ ⎠⎦
Figure A2. The dependence of carrier concentration in the graphene channel on SiO2 substrate on
gate voltage for different sweeping rates of the gate voltage. Forward sweep ( dVg (t ) dt > 0 ) is
4
shown by red curves, backward one ( dVg (t ) dt < 0 ) – by the black ones. Parameters, used for
Figure A.2 presents the dependence of 2D concentration in graphene channel vs. the gate
voltage as a function of the gate voltage switching rate. Under slow sweeping rates the surface
dipoles polarization renews, under fast ones it vanishes completely. This corresponds qualitatively
the experimental data 6 for the graphene channel on SiO2 substrate with the absorbed water molecules
on it’s surface. At that the increase of dVg dt leads to narrowing and later on to vanishing of
Figure A3. The dependence of carrier concentration in the graphene channel on PZT substrate on
gate voltage for different rates of gate voltage sweeps. Forward sweep ( dVg (t ) dt > 0 ) is shown by
red curves, backward one ( dVg (t ) dt < 0 ) – by the black ones. Parameters, used for calculations: κ =
5
Figure A.3 presents the dependence of 2D concentration in graphene channel vs. the gate
voltage as a function of the gate voltage switching rate. Under slow sweeping rates the surface
dipoles polarization renews, under fast ones it vanishes completely.
References
1
Кurchak, A. I.; Morozovska, A.N.; Strikha, М.V. Rival mechanisms of hysteresis in the resistivity of
graphene channel. Ukr. J. Phys. 2013, 58, 473.
2
Das Sarma, S.; Adam, S.; Hwang, E. H.; Rossi, E. Electronic transport in two-dimensional graphene. Rev.
Mod. Phys. 2011, 83, 407 – 470.
3
Strikha, M.V. Mechanism of the antihysteresis behavior of the resistivity of graphene on a Pb(ZrxTi1–x)O3
ferroelectric substrate. JETP Letters 2012, 95, 198-200.
4
Kurchak, A.I.; Strikha, M.V. Antihysteresis of the electrical resistivity of graphene on a ferroelectric
Pb(ZrxTi1– x)O3 substrate. JETP 2013,143, 129–135.
5
Zheng, Y.; Ni, G.-X.; Toh, C.-T.; Tan, C.-Y.; Yao, K.; Ozyilmaz, B. Graphene field-effect transistors with
ferroelectric gating. Phys. Rev. Lett. 2010, 105, 166602.
6
Wang, H.; Wu, Y.; Cong, C.; Shang, J.; Yu, T. Hysteresis of Electronic Transport in Graphene Transistors.
ACS Nano 2010, 4, 7221–7228.