Nonlinear screening in

graphene nanostructures
Michael Fogler, UCSD
 Acknowledgements
Work in collaboration with:
„ L. Glazman (Yale)
„ D. Novikov (Yale)
„ B. Shklovskii (U MN)
„ Matt Zhang (UCSD)
Funding
„ NSF, ACS UCSD
Quasi-relativistic (“Dirac”) fermions

E (p) = ± υ | p |
300 times slower
υ ≈ c / 300 than light

E > 0 "electrons"
E < 0 "positrons" or "holes"

Fermi surface shrinks to a single point

 Graphene: QED in a pencil trace?
Similarities Differences
Linear spectrum 2+1D instead of 3+1D
Spin-like degree of No mass
freedom Interactions are strong
Chirality No retardation (υ < c )
Doping, disorder,
phonons
….
 Klein paradox (1929)
An example of what “QED in a pencil trace” may test
Katsnelson
Naïve calculation gives transmission > 1 ??
„ Source of mistake: for the negative-energy states
(group velocity) = - (quasi-particle velocity)
electron states

e p
0
positron states
The result of the correct calculation: transmission < 1 and
increases with the barrier height
– Not exactly paradoxical in 2007 but still unusual ...
 Supercritical charge: atomic
collapse and creation of antimatter
Energy Schwinger, 1950’s
Zeldovich, 1970’s

2mc 2

Zα < C ~ 1
+Z +Z
subcritical
Zα > C
 Fine-structure constant in graphene
2
e 1
α= ≈
hc 137
2
e
α Graphene = ≈ 0.9
κ hυ
κ = 2.3 dielectric constant
SiO2
High-k dielectric: weak interactions

2
e
α= , κ = dielectric constant
κ hυ

If κ >> 1, α << 1 HfO2, water, …

κ
 Our work – nonlinear
screening in graphene

References: arXiv:0707.1023 (PRB

2007), 0708.0892, and 0710.2150
 What is screening?

+
+ + V (r ) r

+Z +

+
Vext (r )
+
+
 Linear screening
Poor screening ε (q) ≅ 1 Good screening
External charge ~ sin qx ε (q ) >> 1
Induced rs
charge
x x

rs
…
1 Guinea et al.
rs = screening length Ando
α kF Das Sarma et al.
…
Why would screening be nonlinear
in graphene?
Lots of carriers Lots of carriers
good screening No carriers good screening
no screening!
 Nonlinear screening

Higher electron density, Lower density,

better screening worse screening

rs rs
x
Summary of possible screening
regimes
Screening Poor Poor Good

Valid N/A Thomas- Classical

approximation (Dirac eq.) Fermi electrostatics

λF
0 λF rs = Length
α scale

If α << 1
 Graphene p-n junction
Electrostatic potential
of the gates induces
a gradually varying
charge density
profile, which
changes sign.
„ The back-gate
controls the overall
carrier density
„ The top gate
determines the
density difference
 Recent experiments
Resistance of a graphene
p-n junction is only a few
kΩ
Qualitatively explained by
the gapless Dirac
spectrum
Quantitative theory needed

Williams et al., Science 317, 638 (2007)

B. Huard et al., PRL 98, 236803 (2007)

B. Ozyilmaz et al., arXiv/cond-mat:0705.3044
 Klein paradox (1929)
An example of what “QED in a pencil trace” may test
Naïve calculation gives transmission > 1 ??
„ Source of mistake: for the negative-energy states
(group velocity) = - (quasi-particle velocity)

electron states

e p
0
positron states

The result of the correct calculation: transmission < 1 and

increases with the barrier height
– Not exactly paradoxical in 2007 but still unusual ...
 Veselago lensing

Cheianov et al. 2007

 Electric field in the junction
Conductance is determined
by the field strength at the
V
interface
A naïve estimate (Cheianov
06) assumed uniform field:
x
In reality, electric field is
„ away from
suppressed
junction where V
screening is good
„ enhanced near the x
interface where
screening is poor
Nonlinear screening, case α~1
Electron density is highly non-
uniform, how to compute the
screened potential?
It is charge density that must
be found first!
Screened potential V(x) F Field
follows from the density of enhancement
near interface
states
Three-line derivation, next
x
 Nonlinear screening, α~1
Charge density ρ
ρ ( x) = ρ ' x p n

Potential, T-F approx. x

eV ( x) = μ ( x) = hυ ρ ( x) ∝ x
Electric field F
F = −V ' ( x) ∝ 1 / x
Cut the divergence
x* = λF ( x* ) − x* x* x
Numerical results

Junction resistance
h −1/ 6
R = 1.0 2 α ( ρ ' ) −1 / 3

e
 Results
Previous results for transmission through
p-n junction are off by a (parametrically)
large factor (~2-10, in practice)
Our new results bring theory and
experiments in a much better agreement
Including disorder effects the agreement
can be made quantitative (next slide)
Disorder in p-n junction
Coulomb impurity in
graphene
 Why the Coulomb impurity
problem?
Uncontrolled charged impurities in the
substrate
Intentional doping / gating
Intriguing analogy to atomic collapse and
vacuum breakdown in QED

α Graphene ≈ 0.9 > α c ?

 Supercritical charge: atomic
collapse and creation of antimatter
Energy Schwinger, 1950’s
Zeldovich, 1970’s

2mc 2

Zα < 1
+Z +Z
subcritical
Zα > 1
Critical charge in graphene

Z cα ~ 1/ 2

α Graphene ≈ 0.9 ∴ Z c ~ 1 ?

Khalilov (1998)
Novikov
Shytov, Levitov
Castro-Neto
…
 Critical charge in graphene:
previous work
subcritical Z supercritical Z
Electron density Electron density
1
n( r ) ~ α δ ( r ) n( r ) ~ 2 2
r ln r
r r

1 1
V (r ) ~ V~
r r ln r

DiVincenzo, Katsnelson,
Shytov, Castro-Neto
 Problem with previous work: range
of validity
Electron density Short-range cutoff:
bandwidth
1
n( r ) ~ 2 2 Long-range: zero
r ln r
mass
For αZ~1 both
cutoffs ~ lattice
r constant
Our work – “hyper”critical charge
M.F., Novikov, Shklovskii, PRB (2007)

α Z >> 1
 New result for “vacuum
polarization” around a large
Coulomb charge
supercritical Z New universal result
for the structure of
Electron density
1 the supercritical core
n( r ) ~ 3
r For α =1 this core
r “squeezes out” the
regimes discussed in
previous literature
1 Such regimes return
V~
r 3/ 2 if α << 1 (next slide)
Hypercritical impurity, small α
How to realize large-Z charge?

STM / AFM
tip

d +Z