NCN@Purdue Summer School, July 20-24, 2009

Low-Bias Transport in Graphene:

an introduction
Mark Lundstrom and Dionisis Berdebes
Network for Computational Nanotechnology
Discovery Park, Purdue University
West Lafayette, IN

1
 acknowledgments

Yang Sui, Changwook Jeong, Raseong Kim

Tony Low, Supriyo Datta and Joerg Appenzeller

A set of notes to accompany this lecture is available. The notes provide

derivations for all of the equations presented in this lecture, as well as
additional discussions for TL > 0K, the role the the graphene quantum
capacitance, and derivations of scattering rates. See:

D. Berdebes, T. Low, and M.S. Lundstrom, “Lecture notes on Low bias

transport in graphene, July 2009.”

2
 outline

1) Introduction and Objectives

2) Theory
3) Experimental approach
4) Results
5) Discussion
6) Summary

3
 graphene

Graphene is a one-atom-thick planar carbon sheet with a

honeycomb lattice.

source: CNTBands 2.0 on nanoHUB.org

Graphene has an unusual bandstructure that leads to

interesting effects and potentially useful electronic devices.

4
 objectives

• Describe the experimental techniques commonly-used to

characterize low-bias conductance of graphene.

• Show some typical results.

• Analyze the results and discuss the general features of

low-bias transport in graphene and how they are related to
carrier scattering.

5
 outline

1) Introduction and Objectives

2) Theory
3) Experimental approach
4) Results
5) Discussion
6) Summary

6
 simplified graphene bandstructure
We will use a very simple description of the graphene bandstructure,
which is a good approximation near the Fermi level.

E (k )
E (k ) =
± υ F k =
± υ F k x2 + k y2

υ ( k )= υ F ≈ 1×1 80cm/s
gV = 2
f1 (E ) EF
D( E ) = 2 E π  2υ F2
k
“neutral point”
(“Dirac point”) We will refer to the EF > 0 case, as
“n-type graphene” and to the EF < 0
case as “p-type graphene.”
7
 low-bias transport theory

0V W I V

f1 (E ) f2 (E )
graphene
EF 1 EF 2 = EF 1 − qV

I=
2q
+∞

∫ T (E )M (E )( f1 − f2 )dE
f0 ( = (
E ) 1 1 + e(
E − EF ) k B T
)
h −∞

2 +∞
T ( E ) ≡ λ (E ) (λ (E ) + L)
I 2q
G=
V
=
h ∫ T ( E )M ( E ) ( − ∂f 0 ∂E ) dE
−∞ M ( E ) = W 2 E π υ F
8
 expected results: G vs. EF at TL = 0K

G vs. EF
E M (E ) ∝ E

EF > 0

EF = 0
T (EF ) ≈ constant M (E )
EF
∂f (E ) EF < 0
− 0 = δ (E − EF )
∂E

2q 2
G ( 0K ) = T ( EF ) M ( EF )
h
9
 expected results: G vs. nS at TL = 0K

2q 2
G= T ( EF ) M ( EF )
G vs. EF h

2
vs. nS 1  EF 
nS ( EF )
=   ∝ E 2

π  υ F  F

M (EF ) ∝ EF ∝ nS
T (EF ) ≈ constant

nS
G ∝ nS

10
 expected results: TL > 0K

T > 0K G E M (E ) ∝ E
EF > 0

EF = 0

M (E )
T = 0K

EF
EF < 0
GS (EF = 0 ) > 0

2q2
G (TL > 0 K ) = T ( EF ) M ( EF )
h 11
 some key equations (T = 0K)

Describes the conductance of the

2q2
G (0 K ) = T ( EF ) M ( EF ) conduction (E > 0) or valence (E < 0)
h bands.
M ( EF ) = W 2 EF π υ F
(For T > 0, the total conductance is the
sum of the two.)
T (EF ) = λ (EF ) (λ (EF ) + L)

2q 2 λ ( EF ) 2 EF
G ( 0K ) = W
h λ ( EF ) + L π υ F

W Gs is the “sheet conductance” or

G = GS conductivity, σ
L
2q 2  2 EF  1 1 1
GS ( 0K ) = λapp   = +
h  π υ F  λ12app λ (EF ) L
 an aside

When EF > 0, graphene is strongly degenerate and:

 2q 2 k BTL  2q 2  2 EF 
GS ( EF )  2 2  λapp F 0 ( EF k BTL ) ≈ λapp ( EF )  
 π  υ F  h  π υ F 

TL > 0K result ≈ TL = 0K result

2 2
 2   k BTL  1  EF 
nS    1( F B L)
F E k T ≈  
π
  F  υ π 
 Fυ

13
 questions

• How is G vs. EF (or G vs. nS) measured experimentally?

• How do the results compare to theory?

• What do the results us about scattering in graphene?

For a comprehensive review of theoretical aspects of graphene, see: A. H.

Castro Neto, F. Guinea, N. M. R. Peres, Braga, Portugal, K. S. Novoselov, and
A. K. Geim. “The electronic properties of graphene,” Reviews of Modern
Physics, Vol. 81, pp. 109-162, 2009.

14
 outline

1) Introduction and Objectives

2) Theory
3) Experimental approach
4) Results
5) Discussion
6) Summary

15
 gate-modulated conductance in graphene

1) The location of the Fermi level (or equivalently the

carrier density) is experimentally controlled by a “gate.”

2) In a typical experiments, a layer of graphene is place

on a layer of SiO2, which is on a doped silicon
substrate. By changing the potential of the Si substrate
(the “back gate”), the potential in the graphene can be
modulated to vary EF and, therefore, nS.

16
 experimental structure (2-probe)
(4-probe is used to eliminate series resistance and for Hall effect measurements.

graphene
W, L, ~ microns
SiO2
L
Back gate W graphene
(doped Si)

SiO2

Side view Top view

Typically, Cr/Au or Ti/Au are used for the metal contacts.

The thickness of SiO2 is typically 300nm or 90nm, which makes it
17
possible to see a single layer of graphene.
 suspended graphene

graphene
SiO2

Back gate SiO2 removed by etching.

This eliminates charges in the SiO2
(doped Si)
and after annealing produces higher
quality graphene.

“Temperature-Dependent Transport in
Side view
Suspended Graphene”
K. Bolotin, et al., Phys. Rev. Lett. 101,
096802 (2008)

18
 measurements

− V + I
G=I V R=V I
graphene
SiO2
At a fixed temperature:

Back gate G (VG ) or R(VG )

(doped Si)
At a fixed gate voltage:

G (TL ) or R(TL )
VG
Frequently the sheet conductance or sheet resistance is reported
(and this is usually referred to as the ‘conductivity’ or the ‘resistivity.’)
G = GS (W L) 19 R = RS (L W )
using a gate voltage to change the Dirac point (or EF)
∆V I E (k )

graphene
SiO2
VG′ > 0

Back gate
VG′ = 0
(doped Si) k

VG′ < 0

VG

VG′ = VG − VNP
20
 gate voltage - carrier density relation

nS (pS )
VG′ = VG − VNP VG

If the oxide is not too thin (so that

the quantum capacitance of the
graphene is not important), then:

qnS = CinsVG′

VNP VG (VG′ ) ε ins

Cins =
tins

21
 outline

1) Introduction and Objectives

2) Theory
3) Experimental approach
4) Results
5) Discussion
6) Summary

22
 sheet conductance vs. VG

G = GS W L
1 µm

2q 2  2 EF 
GS ( EF ) ≈ λapp ( EF )  
↑ h  π υ F 
L ≈ 5000 nm
GS t ox = 300 nm 2
1  EF 
mS =
nS CoxVG ≈  
B= 0 T = 10 K π  υ F 

VG (V ) → λapp ( EF ) =
GS ( 2q h )
2

2 nS π
Fig. 30 in A. H. Castro, et al.,“The electronic
properties of graphene,” Rev. of Mod. Phys.,
81, 109, 2009.

23 (T L
=0K )
 mean-free-path (VG = 100V)

1 µm GS ≈ 3.0 mS

nS ≈ 7.1 × 1012 cm -2
↑
L ≈ 5000 nm
GS t ox = 300 nm E F ≈ 0.3 eV
mS
B= 0 T = 10 K
λapp (0.3 eV )≈ 130 nm
VG (V ) →

λ (0.3 eV )<< L
Fig. 30 in A. H. Castro, et al.,“The
electronic properties of graphene,”
Rev. of Mod. Phys., 81, 109, 2009.
24
 mean-free-path (VG = 50V)

GS ≈ 1.5 mS
1 µm
nS ≈ 3.6 × 1012 cm -2

↑ E F ≈ 0.2 eV
L ≈ 5000 nm
GS t ox = 300 nm
mS λapp (0.2 eV )≈ 90 nm
B= 0 T = 10 K
λ (0.2 eV )
≈ 0.69
VG (V ) →
λ (0.3 eV )

0.2 eV
Fig. 30 in A. H. Castro, et al.,“The ≈ 0.67
electronic properties of graphene,” 0.3 eV

λ (E F )∝ E F
Rev. of Mod. Phys., 81, 109, 2009.
25
 mobility

1 µm Since, GS ~ nS, we can write:

GS ≡ nS qµ n
↑
GS
and deduce a mobility:
mS
B= 0 T = 10 K
µ n ≈ 12,500 cm 2 /V-sec
VG (V ) →
Mobility is constant, but mean-
free-path depends on the Fermi
Fig. 30 in A. H. Castro, et al.,“The energy (or nS).
electronic properties of graphene,”
Rev. of Mod. Phys., 81, 109, 2009.
26
 VG = 0

1 µm

GS ≈ 0.16 mS
↑
L ≈ 5000 nm
GS t ox = 300 nm nS = CoxVG ≈ 0 ?
mS
B= 0 T = 10 K GS
λapp =
( 2q 2 h ) 2 nS π
VG (V ) →

λapp → ∞ ?
Fig. 30 in A. H. Castro, et al.,“The
electronic properties of graphene,”

(T )
Rev. of Mod. Phys., 81, 109, 2009.
=0K
27 L
 electron-hole puddles

GS 1 µm
1/kΩ

B= 0 T = 10 K

VG (V ) →

Fig. 30 in A. H. Castro, et al.,“The electronic J. Martin, et al, “Observation of

properties of graphene,” Rev. of Mod. electron–hole puddles in graphene
Phys., 81, 109, 2009. using a scanning single-electron
transistor,” Nature Phys., 4, 144, 2008
28
 effect of potassium doping

For nominally undoped

samples, GS vs. nS is non-
linear.

↑
GS
 2q 2  As doping increases, GS
  vs. nS becomes more linear,
 h 
mobility decreases, and the
NP shifts to the left.

VG (V ) →

J.-H. Chen, C. Jang, S. Adam, M. S. Fuhrer, E. D. Williams,and M. Ishigami,

“Charged-impurity scattering in graphene,” 29
Nature Phys., 4, 377-381, 2008.
 nominally undoped sample

t ox = 300 nm

λ << L
λapp =
GS (2q h)≈ 164 nm
2

T = 20 K
2 nS π

λ << L

VG (V ) →

J.-H. Chen, C. Jang, S. Adam, M. S. Fuhrer, E. D. Williams,and M. Ishigami,

“Charged-impurity scattering in graphene,” 30
Nature Phys., 4, 377-381, 2008.
 temperature dependence
Away from the conductance minimum, the conductance decreases as TL
increases (or resistivity increases as temperature increases).

TL < 100 K : RS ∝ TL
(acoustic phonon scattering - intrinsic)

hω
T
 L > 100 K : RS ∝ e 0 kBTL

(optical phonons in graphene or

surface phonons at SiO2 substrate)

J.-H. Chen, J. Chuan, X. Shudong, M. Ishigami, and M.S. Fuhrer, “Intrinsic and extrinsic
performance limits of graphene devices on SiO2,” Nature Nanotechnology, 3, pp. 206-209,
2008. 31
 phonons and temperature dependence

1 1 1
RS = ∝ ∝ Nβ Nβ =
GS λ e ω ( β ) kBTL − 1

acoustic phonons: optical phonons:

ω < kBTL ω 0 ≈ kBTL

k BTL 1
Nβ ≈ Nβ =
ω e ω0 k BTL
−1
1
RS ∝ TL RS ∝
e ω0 k BTL
−1

32
 unannealed vs. annealed suspended graphene

GS ∝ nS

λapp ≈ 1300 nm

expected from
ballistic theory

T = 40 K

GS = nS qµ n + Gres
K. I. Bolotin, K. J. Sikes, J. Hone, H. L. Stormer, and
P. Kim, “Temperature dependent transport in
suspended graphene,” 2008 33
 about mobility

2q 2  2 EF 
GS ( EF ) ≈ λapp ( EF )  
h  π υ F 

( )
GS (E F ) ∝ λapp E F nS
Case 1): Case 2):
GS ≡ nS qµ n
λapp ∝ E F ∝ nS λapp constant

λapp (E F ) GS ∝ nS GS ∝ nS
µn ∝
nS
µ n constant µn ∝ 1 nS

34
 experimental summary: graphene on SiO2

1) Low conductance samples often show GS ~ nS (away from the

minimum)

2) Higher conductance samples are frequently non-linear (GS rolls off at

higher nS)

3) GS(T) decreases with temperature (“metallic”) for large nS

4) RS ~ TL for TL < 100K and superlinear for TL > 100K

5) Best mobilities for graphene on SiO2 are ~30,000 cm2/V-s at TL = 5K

6) Asymmetries between +VG and -VG are often seen.

35
 experimental summary: suspended graphene

1) Before annealing GS ~ nS (away from the minimum)

2) After annealing, GS increases and GS vs. nS becomes non-linear

3) After annealing, GS is close to the ballistic limit

4) Best mean-free-paths are ~ 1μm at TL = 5K

5) GS decreases with TL for large nS but increases with TL near the

Dirac point.

36
 outline

1) Introduction and Objectives

2) Theory
3) Experimental approach
4) Results
5) Discussion
6) Summary

37
 conductance and scattering

2q 2 λ ( EF ) 2 EF
G ( 0K ) = W
h λ ( EF ) + L π υ F

λ(E) is the mean-free-path (technically, the mfp for “backscattering”),

which is determined by the dominant scattering processes.

38
 scattering


k′, E ′ 1 scattering rate per sec
τ (E )
typically computed from
FGR

k, E

y US ( ρ )

λ (E ) ∝ υ F τ (E )
x mean-free-path for backscattering

39
 scattering

π
λ (E ) = υ F τ m (E ) (elastic or isotropic scattering)
2

For many scattering mechanisms (e.g. acoustic phonon, point defect),

the scattering rate is proportional to the density of final states:

1
∝ D (E ) ∝ E τ (E ) ∝ E −1
τ (E )

The energy-dependent mean-free-path is:

λ (E ) ∝ 1 E

What does this type of scattering do to the conductance?

40
 effect of short range / ADP scattering

Assume TL = 0 K and diffusive transport (just to keep the math simple)

2q 2  2 EF 
GS = λ ( EF )   λ ( EF ) ∝ 1 EF
h  π υ F 

GS = constant!

For short range or ADP scattering, GS is constant.

N.H. Shon and T. Ando, J. Phys. Soc. Japan, 67, 2421, 1998.

41
 long range (charged impurity) scattering

E λ ↑ as E ↑
W, L, ~ microns

graphene
ENP (x) = U S (x)
x
SiO2

Top view Random charges introduce random

For screened or unscreened
charged impurity scattering, the
mfp is proportional to energy.
42
fluctuations in E(k), which act a
scattering centers.
High energy electrons don’t “see”
these fluctuations and are not
scattered as strongly.
 effect of charged impurity scattering

Assume TL = 0 K and diffusive transport (just to keep the math simple)

2q 2  2 EF 
GS = λ ( EF )   λ (EF ) ∝ EF
h  π υ F 

GS ∝ nS (µn constant )

For charged impurity scattering, GS vs. nS is linear.

T. Ando, J. Phys. Soc. Japan, 75, 074716, 2006

N.M.R. Peres, J.M.B. Lopes dos Santos, and T. Stauber,
Phys. Rev. B, 76, 073412, 2007.
43
 comment on linear G vs. nS

The observation of a linear G(nS) characteristic is frequently

taken as experimental evidence of charged impurity scattering,
but…

Theoretical work shows that strong, neutral defect scatter can

lead to a linear G vs. nS characteristics…

T. Stauber, N.M.R. Peres, and F. Guinea, “Electronic transport in

graphene: A semiclassical approach including midgap states,”
Phys. Rev. B, 76, 205423, 2007.

Even more recent experimental work on intentionally damaged

graphene bears this out…

TJ.-H. Chen, W.G. Callen, C. Jang, M.S. Fuhrer, and E.D.

Williams, “Defect Scattering in graphene,” Phys. Rev. Lett., 102,
236805, 2009.
 the energy-dependent mfp

Mobility is not always the best way to characterize the quality of a

graphene film, but mean-free-path is always a well-defined quantity.
We can extract the mean-free path vs. energy from measured data.

2
 2 EF  GS (Vg ) ( 2q 2 h )
GS ( 0K ) =
2q
λapp ( EF )   λapp ( EF ) =
h  π υ F  2 nS (Vg ) π
2
1  EF 
nS ( 0K ) =   =
1 1
+
1
π  υ F  λapp ( EF ) λ ( EF ) L

The apparent mfp is the shorter of the

actual mfp and the sample length.
45
 example

TL = 40K

“Temperature-Dependent Transport in
Suspended Graphene”
K. Bolotin, et al., Phys. Rev. Lett. 101,
096802 (2008)

46
 suspended, annealed

TL = 40K

“Temperature-Dependent Transport in
Suspended Graphene”
apparent mfp apparent mfp
K. Bolotin, et al., Phys. Rev. Lett. 101, increases with independent of
096802 (2008) energy energy -
approximately L
47
 suspended, annealed

GS (Vg ) (2 q 2 h ) λapp (EF )

λapp (EF ) =
( )
2 nS Vg π

1 1 1 λ (EF )
= +
λapp (EF ) λ (EF ) L

(data from: K. Bolotin, et al., Phys. Rev. Lett. 101, 096802, 2008)
48
 suspended, unannealed
linear GS vs. n suggests charged
impurity scattering.

Expect ~λ |E|

T = 40 K

K. Bolotin, et al., PRL 101, 096802 (2008)

analysis complicated by
large residual resistance.
49
minimum and residual conductance

Gres ≈ Gmin
Gmin

T = 40 K

K. Bolotin, et al., PRL 101, 096802 (2008)

q2
Gres ≈ 14 G (nS ) = Gres + (qµ1 )nS
h
50
 suspended, unannealed

 ~ |E|
as expected

GS (Vg ) − Gres  ( 2q 2 h )

λapp ( EF ) =  
2 nS (Vg ) π

51
 general picture of GS vs. nS (ballistic)

ballistic
G ∝ nS

If the mfp is small and constant, then

G is also proportional to sqrt(nS), but
the magnitude is less than the
ballistic limit.

nS

We have discussed Vg (nS) > 0, but by symmetry, the same thing

should occur for p-type graphene (EF < 0).

52
 general picture of GS vs. nS (diffusive)
Charged-impurity
Short range or ballistic (long-range)
acoustic phonon G scattering.
scattering. ∝ nS
Constant mobility.

Result is a combination
of charged impurity and
phonon scattering.

nS
Non-zero residual resistance
commonly observed.

53
 outline

1) Introduction and Objectives

2) Theory
3) Experimental approach
4) Results
5) Discussion
6) Summary

54
 summary

• The general features of the graphene conductance vs. gate voltage

are readily understood (but still being discussed).
• Data can be analyzed by extracting the mean-free-path for
backscattering and relating it to the underlying scattering
mechanisms.
• More sophisticated theoretical treatments include screening, remote,
polar phonons, etc.
• Actual experiments are frequently non-ideal (e.g. not symmetrical
about VNP, non monotonic behavior, variations due to sample state,
uncertainties in W and L, etc.
• But the material presented here gives a general framework and
starting point for analyzing experimental data.

55
what about the contacts?

− V + I

graphene
SiO2

Back gate
(doped Si)

VG

56
 suspended, unannealed
linear GS vs. n suggests charged
impurity scattering.

Expect ~λ |E|

T = 40 K

K. Bolotin, et al., PRL 101, 096802 (2008)

analysis complicated by
large residual resistance.
57