Selective-Area Atomic Layer

Deposition of Copper

Nanostructures for Direct Electro-

Optical Solar Energy Conversion

Brian Willis

UCONN

Cancun, MX 2014

Plasmonics

Plasmonic nanoparticles have applications for localized heating,

SERS, nanophotonics, catalysis, sensors and more…

Au, Ag, Cu Nanoparticles interact strongly with visible/ near IR

radiation, tunable by size/shape.

plasmons=collective modes that strongly

concentrate light at the nanoscale.

+ + + +

- - - -

E

Quasi-Static Approximation

𝛼 = 4𝜋𝑎3𝜀(𝜔) − 𝜀𝑚𝜀(𝜔) + 2𝜀𝑚

Resonance at Re [𝜀(𝜔)] = -2𝜀𝑚

Van Duyne, Nat. Mater. 2008

Plasmonic Dimers

Nordlander, Science 2014

Geometric effect concentrates electric fields at nano-gaps

leading to large field enhancements.

EM field induces a voltage drop across the nanoelectrodes.

“hot spots” are created by “plasmonic

dimers”

Plasmonic Enhancement

Mayer, J. Phys. Conds. Mater. 2009

Theoretical studies predict large enhancements in rectification

ratios for plasmonic resonances.

Lightning rod effect concentrates field at tip = asymmetry.

0.5 nm radius tip

Ag tip

/ is rectification ratio

Geometric Asymmetry

lightning rod

effect

Optical Rectification

Concept: direct conversion of EM radiation (solar) into DC

power by antenna – coupled high speed diodes to rectify

optical frequency charge waves.

Potential Advantages: low cost fabrication, tunable

absorbance including IR (waste heat), and device integration.

2

2

2

4

1)()(

V

IVVIIVII

photoDCphotoDC

Antenna

DC Filter Load

Diode

No band-gap limitations!

Literature Data

Ward et al have demonstrated optical

rectification in an electromigration

junction.

- not tuned for plasmonic resonance

- not scalable

- no control of geometry

- isolated nanostructures

Ward, Nat. Nano. v. 5, 732 (2010)

Rectenna Concept

Electrically connected

Sized for plasmon

resonance in visible

Asymmetric geometry

Nanoscale tunnel gap

Not commercially available!

Rectenna Challenges

Nanoscale antenna tuned to visible/near-IR

Geometric asymmetry, diode response at low voltage

Electrically contacted, tunneling devices

RC time constant, impedance matching

Extremely fast diodes, 1015 Hz

Diode is most critical need!

Rectenna Concept

Electrically connected

Asymmetric geometry

Sized for plasmon

resonance in visible

Nanoscale tunnel gap

ALD used to make tunnel diodes

ALD Metals Review

Choices for ALD Cu, Ag, and Au are limited, non-ideal.

Cu(II)thd2/H2 process is well-behaved for selective growth.

Year Reactants GPC

(nm)

T window

(°C) type

1998 Cu(II)thd2 // H2 0.03-0.04

190-260 thermal

2003 bis(N,N’-diisopropylacetamidinato)diCu(I) // H2 0.01-

0.05 220-300 thermal

2013 1,3-Diisopropyl-imidazolin-2-ylidene Cu(I) hexamethyldisilazide

// H radicals 0.02 250 plasma

2014 Cu(II)(OCHMeCH2NMe3)2 // BH3(NHMe2) 0.013 130-160 thermal

2007 (2,2-dimethylpropionato) Ag(I)triethylphosphine // H radicals 0.12 140 plasma 2010 (hfac)Ag(I)(1,5-COD) // propanol - 110-150 thermal

2011 Ag(I)(fod)(PEt3) // H radicals 0.03 120-150 plasma

2014 (hfac)Ag(I)(PEt3) // HCHO - 170-200 thermal

Au??

Selective Area Growth

oxide/nitride

metal

seed No growth

Growth

Selective area ALD enables deposition on seeded regions

and eliminates the need for etching.

SA-ALD enables nanostructures.

Plasmonic Structures

10x10 “Rectangular Dipole”

Rectenna Array, 550 nm x 800

nm spacing

10x10 “Large Triangle” Rectenna Array

500 nm x 500 nm spacing

Electrically isolated structures are used to investigate tuning

of plasmonic response by selective area ALD.

Pd nanostructures

Resonance Tuning

FDTD simulations predict that resonance red-shifts for

decreasing gap size and for increasing nanostructure size.

Gap size dependence Nanostructure size dependence

bar-shape structures

bar size: 150 x 50 x 35 nm

Tuning Plasmonics with ALD

Plasmonic dimers have strong electric field enhancement

>103, with fields > 5x108 V/m 1.

“lightening rod” effect further

intensifies electric fields

1Ward et al., Nature Nano v. 5 p. 732 (2010)

FDTD

simulations

Optical measurements

Focus spot is 3-7 um; sized to the test array.

Optical Response

Spectroscopy data show red-shift after ALD.

Red-shift &

increased

extinction

Red-Shifts

Red-shift magnitude tracks ALD thickness/cycles.

In-situ measurements would provide feedback on nanogap.

Nanoscale Tunnel Junctions by ALD

geometric

asymmetry

Device arrays with 1000’s of tips converge to tunneling similarly

to suspended structures.

Geometric asymmetry contributes to desired diode IV character.

SiO2

0

1 10-8

2 10-8

3 10-8

4 10-8

5 10-8

0 200 400 600 800 1000

Curr

ent

(A)

Time (seconds)

-0.4 -0.2 0 0.2 0.4

d2I/d

V2 In

ten

sity

Volts (V)

Trapping and IETS detection of Acetic Acid

molecules in nanojunctions

adsorption

IETS

ALD Grown Tunnel Junctions

HIM Images

< 9 nm < 7 nm

HIM provides some estimates for gap size.

Hang Dong, Rutgers

Nano-Nucleation Challenges

Growth is sensitive to surface pre-treatment, growth inhibition is

observed without UVO.

ALD metal films are not layer-by-layer, Nanograins form with inherent

surface roughness. How smooth can we achieve? Could be shape

dependence, crystal structure effects.

Without UVO With UVO

UV/Ozone Sample Pre-treatment

planar films

ALD Cu roughness values are similar to

other reports2, scales with thickness.

UVO enhances/enables growth, removes most C 1s, but residual C

remains even after long time (60 min).

acetate-like

2Winter et al. Chem. Mater. v. 26, p. 3731 (2014)

XPS

Selectivity

Selectivity is lost at higher temperatures, inverted selectivity?

Selectivity was a mystery; how does Cu grow on oxide?

Annealed in He at 600°C prior to

growth at 220°C High T growth, 230°C

ALD Fundamentals with in-situ SE

Cu(thd)2 is a solid source precursor. H2 is co-

reactant.

In-situ, real time SE provides saturation curves.

note O

content of

precursor

SE Growth Trajectory

in-situ SE provides growth “finger-print”, sensitive to

substrate preparation and growth conditions.

extracted GPC ~ 0.04 nm/cycle matches SEM data.

~15/466 = 0.03

delta/cycle

Original Mechanism

Cu(thd)2 adsorbs to saturation during precursor dose.

Martensson et al, JES, v. 145, p. 2926 (1998)

Martensson’s

mechanism

SE Growth Signature

SrO ALD

SrO ALD shows characteristic signature for stable precursors

adsorption.

Cu ALD has opposite signature with reversible adsorption.

Cu ALD

Original Mechanism

Selectivity follows from H2 dissociation requirement.

Original mechanism is NOT consistent with CuI CVD by

disproportionation at 150-200°C, and can’t explain

observed GPC. Martensson et al, JES, v. 145, p. 2926 (1998)

Martensson’s

mechanism

New Mechanism

New mechanism explains SE signature and is consistent

with reversible adsorption. Also explains high GPC.

Roll of Pd-H2 is now more complex, strong chemisorption.

SE Growth vs. Temperature

Mechanism not sensitive to temperature. H* must be stable.

Dissociative reversible adsorption is consistent with

thermodynamics if Hads ~ 34 kcal/mol.

Real-time SE vs. Temp Coverage Simulations

CuL2 + 2* CuL* + L*

QCM Studies

QCM measures mass and thermal effects.

Reaction signal (at 150°C) is consistent with RTSE data,

GPC ~ 0.04 nm/cycle.

Role of Pd

Temperature and time dependence of Cu XRD and XPS signals indicate

extensive Cu/Pd mixing. Lower temperatures = higher Cu/Pd ratios.

ARXPS shows evidence for Pd enrichment at surface.

Pd explains H chemisorption. What happens if the Pd runs out?

XRD XPS

Pd segregation

Cu/Pd vs. Cu/Pt

Both Pt & Pd act as seed layers, but Pd mixes more

extensively with Cu.

GPC on Pd/H*

GPC decreases as Cu/Pd becomes more Cu rich.

Cu/Pd ratio measured as 45:1 for this sample.

Cu ALD on Pd

150C

Selectivity

Selectivity is evident from in-situ SE data, but how to explain

non-selective growth on SiO2?

No adsorption/

desorption cycles

for SiO2, Si3N4

Martensson et al, JES, v. 145, p. 2926 (1998)

Non-Selective Growth Mechanism?

C 1s signal is removed by sputtering, but O 1s signal remains. Cu 2p signal

could be Cu0 or CuI. Cu/O ratios are ~1.3 . AES peaks indicates CuI.

CuO/Cu2O growth explains non-selective growth mechanism. Mechanism

requires self-decomposition. H2 not required, but may affect thickness.

For thick Cu films, self-decomposition may contribute to growth.

C 1s Cu 2p O1s

Growth on SiO2 at 205°C

H2 effect on Non-selective Growth

0

0.5

1

1.5

2

2.5

3

3.5

130 140 150 160 170 180 190 200 210

ExcellentGoodPoor

pH

2 (

To

rr)

Temperature (C)160 C

3 Torr

200 C

1.5

Torr

H2 pressure plays a significant role for

non-selective growth. Why?

H2 effect on Non-Selective Growth

180 C

3 torr H2

180 C

1.5 torr H2

180 C

0.5 torr H2

H2 Effect on Selective Growth

0.0 0.5 1.0 1.5 2.0 2.5 3.0

0.000

0.005

0.010

0.015

0.020

0.025

0.030

0.035

GP

C(n

m/c

yc

le)

H2 partial pressure (torr)

140C

160C

GPC Measured by QCM in selective window shows no effect

of H2 partial pressure.

No growth without H2.

Cu ALD on Pd Seed Layers

Summary

ALD Cu is useful for tuning the optical

response of plasmonic materials. ALD

of Ag, Au would be nice.

Selective Area Cu ALD works well and is

useful for tunnel junctions with many

potential applications.

Cu ALD mechanism is partially

understood, Pd, H effects are important.

Acknowledgements

Co-Pi’s (Physics, Penn State): Darin Zimmerman,

Gary Weisel, James Chen

Students: X. Jiang & J. Qi (UCONN)

R. Wamboldt (Penn State)

Funding: NSF-EECS