Two dimensional electron gas

at the SrTiO3 surface

and interfaces

U. Scotti di UccioS. Amoruso, C. Aruta, R. Bruzzese, G. M. De Luca, R. Di Capua, E. Di Gennaro, D. Maccariello, P. Perna, M. Radovic, Muhammad Riaz, Z. Ristic, M. Salluzzo, A. Sambri, R. Vaglio, X. Wang, I. Maggio-Aprile*, and F. Miletto Granozio

CNR-SPIN & University FEDERICO II, Napoli (Italy)

C. Cantoni, J. Gazquez, A. R. Lupini, M. P. Oxley, S. J. Pennycook, N. Plumb, M. Varela

Oak Ridge National Laboratory

*Dép. de Physique de la Matière Condensée, University of Geneva

1

This talk regards two dimensional electron gases that are formed at the STO surface and

interfaces. I will actually present a few data regarding different experiments and

involving several people at CNR SPIN and University of Napoli with the cooperation of a

group at Oak Ridge National Laboratory, mainly involved in TEM measurements.

donors

Polar-non polar interfaces

Where do electrons come from?

Cation

intermixing

Oxygen

vacancies

Donor defects

Electronic reconstruction

LAO

0 0 +1 -1 +1 -1

Po = −−−− ½ e/S

σb = ½ e/S

+

+

+

+

STO

VB

CB

VB

e-

+++++++

-------STO LAO STO LAO

2

This talk:focus on oxygen vacancies

2

A 2DEG is basically characterized by a quantum well where electrons

coming from some donor state are trapped. One of the still debated

issues regarding STO/LAO interfaces is just the nature of the donor

states. In the electronic reconstruction models, electrons come from

the valence band of the polar layer, that is lifted above the Fermi level

by the built-in potential due to the polarization. But one may also

consider defects as donors. These can be due to disorder, but in this

talk I will focus on the effect of oxygen vacancies.

2DEG at STO interfaces

Polar-non polar heterostructures

10-6 mbar Oxygen vacancies dominate!

A. Brinkman, et al., Nature Mat. 2007

3

Reversing the argument:

Can we demonstrate the formation of a 2DEG in

STO as exclusively due to oxygen vacancies?

Can we present interfaces where the role of

oxygen vacancies appears as negligible?

Oxygen vacancies can actually dominate the transport properties of samples that are

fabricated at too low oxygen partial pressure. In that case we can even get bulk

conductivity.

2DEG at STO surfaces and interfaces

STO surface

ARPESon cleaved crystals

Deliberately introduce oxygen vacanciesInvestigate STO surface

Nature 469,189

(2010)

In order to verify whether oxygen vacancies can determine 2DEG formation, we

performed an experiment regarding the free surface of STO. The motivation stems from

a recent work of ARPES on cleaved STO crystals, which indicated that a robust 2DEG is

formed with properties independent on bulk carrier density. I will compare the results

with data concerning oxygen vacancies at the interfaces.

Single crystal (100) SrTiO3 with TiO2 termination planeSelective ethcing procedureUn. Twente, APL 73, 2920 (1998)

AFM

1 µµµµm ×××× 1 µµµµm

STM

2DEG at STO surfaces

after low temperature UHV treatments

5R. Di Capua, M. Radovic, G. M. De Luca, I. Maggio-Aprile, F. Miletto Granozio, N. C. Plumb, Z. Ristic, U. Scotti di Uccio, R. Vaglio, M. Salluzzo

Submitted to Phys. Rev. B

We started with TiO2 terminated STO crystals…

(-10)

(01)

(0-1)

(10)

(-10) (10)

(01)

(0-1)

Thermal treatment

Annealing in oxygen at 850°C

Highly insulatingSurface charging: no LEED pattern

Annealing in UHV at 250-350°C

Bulk insulating (transparent)1×1 reconstruction

Annealing in UHV at 900°CBulk conducting (shiny black)2×1 reconstruction

Sample A

Sample B

6Po = 5×10-11 mbar

…and performed a series of thermal treatments.

Scanning Tunneling

Spectroscopy

Regulation points

V = -1.5 V, I = 0.1 nA

V = -1.5 V, I = 0.7 nA

Average on grid of points

7

tip STO

tip STO

STO surface

STM tip

UHV

Agreement with STS Surf. Sci. 542, 177 (2003)

Agreement with ARPES Nature 469, 189 (2010)

Here is the main result. The dI/dV curves of the different samples show a remarkable

difference. I will stay at the phenomenological level, by saying that the red curve is

consistent with the idea that a high temperature treatment determines conducting bulk

and insulating surfaces, consistently with several previous reports. The blue curve is

instead consistent with the idea that the surface itself is conducting, that is, it possesses

a finite density of occupied states at the Fermi level.

XPS: Ti 3p doublet

Results on 2DEGSpectroscopic signature of 2DEG: Ti 3+ states

Confinement: nm scale

2DEG at STO surfaces

8

Photoelectrons mfp≈nmShallow angle emission: surface probe

XPS

250 °C 900 °C

The photoemission from the Ti3p triplet essentially confirms the existence of occupied

Ti3+ states, that constitute the conduction band, and since the experiment is surface-

sensitive, it also demonstrate the surface confinement within the nm scale.

Since we have no other contamination, the evolution of oxygen is the only possible

source of doping.

VB

Left side of dI/dV:

VB of STOi.e. defect states

9

The investigation of the VB by photoemission also clarifies the transport mechanism.

The left part of the dI/dV plots mainly carry information on the occupied DOS of STO. It

is suggestive to compare between XPS spectra and dI/dV of the different samples; the

data indicate that the surface defect states (blue curve) are source of enhanced junction

conductance.

Out of equilibrium:

flux of vacancies through

the surface into the bulk

Equilibrium:

cs > cb

Oxygen Vacancies at STO surfaces

10

Message

Local doping by Oxygen vacancies

CAN

determine a 2DEG at STO surface

STO environment

Gib

bs

en

erg

y o

f va

ca

nc

ies

First-principles calculationsPHYSICAL REVIEW B 73, 064106 (2006)

The results are consistent with a picture where the Gibbs energy of oxygen vacancies at

the surface is lower, as also deduced by computation. When we raise the chemical

potential of the environment, the vacancies are introduced and at the equilibrium they

reach a higher surface concentration. Finally, if we choose the right conditions, we can

actually form a 2DEG due to local doping.

LAO

STO

LGO

STO

2DEG at STO interfaces

Polar-non polar heterostructures

P. Perna, et al, APL 2010

11Lanthanum Gallate ≈ Lanthanum Aluminate

Now we turn to the issue of 2DEG formation at interfaces STO interfaces. I will consider

on the same foot both STO/LAO and STO/LGO, that in our experience share most

relevant physical properties.

Spectroscopy with atomic-scale resolution

EELS in the aberration corrected microscope

Conducting sampleAbrupt interface

C. Cantoni, et al. ADV. MAT. 2012 12

If we want to investigate the oxygen content at interfaces we need a tool that is

sensitive to chemical properties of samples. We resorted to EELS.

Oxygen vacancy

stoichiometry

δδδδ =0 @1×10-3 mbarbut

2% maximum error 0.12 e- /cell

Integrated O-K

O-K ELNES

C. Cantoni, et al. ADV. MAT. 2012

13

Simulation

Inte

ns

ity (

a.u

.)

experiment

Here are some results. In the upright panel, the integrated O K scan profile confirms the

sensitivity of EELS to determine oxygen stoichiometry with spatial resolution at the

atomic level. However, to get the quantitative information it is necessary to consider

with due care the O-K ELNES spectra, since not all the components are directly

connected to oxygen stoichiometry. Here I show the depth profile of the “c” line

intensity, which is constant across the sampled region within STO.

By quantitatively analyzing the data, we get the best estimation of vacancies

stoichiometry d = 0. This fact is a support to the idea that in a conductive sample grown

at 1x10-3 mbar the 2DEG is not populated by electrons provided by oxygen vacancies.

However, the experimental maximum error is 2%. This means that the required

information is very close to the experimental sensitivity.

We then need further confirmation to conclude that the 2DEG at interfaces can be

populated even in absence of oxygen vacancies.

Further limiting the role of oxygen vacancies

Deposition at high oxygen pressure

DrawbackThe plasma plume is stopped by the buffer gas

The energy of impinging species is reduced

The surface mobility of adatoms drops

growth conditions may be bad!

Energetic particles can favor 2D growth of

oxide thin films through an island break-up

mechanism with prompt insertion, at very low

coverage, and enhanced surface diffusion,

above 50% monolayer coverage

14

To this aim we considered growth at very high oxygen pressure, in spite of the possible

drawbacks.

LGO/STO

Growth at 10-1 mbar

AFM

A. Aruta, et al, APL 2010

Plume analysis and film growth @10-1 mbar

Insulating samples!

Time Resolved Emission Spectroscopy

15

2D growth

Plume fast photography

In our previous work we demonstrated that 2D growth of LAO and of LGO can be

achieved even at 10-1 mbar, but samples were insulating.

More on growth @10-1 mbar

• Shock wave regime• Halting @35 mm• Internal energy drops @30 mm

Interesting regime

Low K (≈eV)High U

High µµµµ(O2)

16

We decided to analyze the growth dynamics by resorting to plume diagnostics such as

fast photography and time resolved spectroscopy. Here are some results.

At 10-1 mbar the plume propagates in a shock wave regime and it halts after travelling

about 35 mm away from the target. Effects of the buffer gas are: the species loose

energy by collisions and reach the substrate with low kinetic energy; instead, they

possess high internal energy; moreover, the oxygen supersaturation is increased by the

strong compression of the gas on the plume front edge. However, care must be paid to

the target-substrate distance.

More on growth @10-1 mbar

R (

Ω

Ω

Ω

Ω /

)

17

By optimizing the distance, conductive samples can be realized. At the given conditions,

it is generally agreed that the content of oxygen vacancies in samples is marginal.

Conclusions

2. STO interfaces

10-3 mbar : V(O) = 0 (EELS)

10-1 mbar: conductive interfaces

1. Oxygen vacancies can determine the formation of a 2DEG at theSTO free surface

Local V(O) 2DEG

18

2DEG Local V(O)

In conclusion, our measurement demonstrate that oxygen vacancies can be collected at

the free surface of STO crystals and there determine the formation of a 2DEG. In the

case of interfaces, on one hand we know that oxygen vacancies can affect the

conductivity; but on the others, we proved that at standard deposition conditions they

don’t accumulate at the interface; and that it is possible to grow conducting interfaces

in extremely oxidizing conditions. In other words, observation of a 2DEG doesn’t seem

to imply the presence of oxygen vacancies.

Conclusions

2. STO interfaces

10-3 mbar : V(O) = 0 (EELS)

10-1 mbar: conductive interfaces

1. Oxygen vacancies can determine the formation of a 2DEG at theSTO free surface

Local V(O) 2DEG

19

2DEG Local V(O)

“different forms of electron confinement at the

surface of SrTiO3 lead to essentially the same 2DEG”

Our measurement demonstrate that oxygen vacancies can be collected at the free

surface of STO crystals and there determine the formation of a 2DEG. In the case of

interfaces, on one hand we know that oxygen vacancies can affect the conductivity; but

on the others, we proved that at standard deposition conditions they don’t accumulate

at the interface; and that it is possible to grow conducting interfaces in extremely

oxidizing conditions. In other words, observation of a 2DEG doesn’t seem to imply the

presence of oxygen vacancies.

In conclusion, it is apparent that different forms of electron confinement (even in terms

of different donor states) at the surface of SrTiO3 lead to essentially the same 2DEG.

MODA LAB: an integrated laboratory

for fabrication of oxide films and for surface analyses

20