Fall 2011

Metal-optic and plasmonic nanolasers:

Current progress and future challenges

Amit M. Lakhani

Advisor: Ming Wu

April 5, 2012

E3S Teleseminar

Fall 2011

Acknowledgements

• Integrated Photonics

Lab

• NSF Graduate

Fellowship

• Funding

– DARPA NACHOS

– NSF E3S

– NSF CIAN ERC

– Samsung GRO

Fall 2011

Outline

• Motivation

• Previous work

• The nanopatch laser

– Description and operation

– Fabrication challenges

• Conclusion

Fall 2011

Interconnect Energy Limits Chip

Performance • 50% of microprocessor power

was in interconnects in 2002

– Expected to rise to >80%

• Current electrical interconnect

– Off-chip ~ 1 pJ/bit

– On-chip ~ 100 fJ/bit

– Wire capacitance

• ~2 pF/cm or 200 aF/mm

• Nanophotonics can lower

interconnect energy

1

2

3

4

Signaling

Wires

(27%)

Clock

(28%)

Logic

(27%)

Memory

(27%)

http://www.itrs.net/Links/2007ITRS/Home2007.htm

Microprocessor Power

Consumption (ITRS 2007)

Fall 2011

Interconnects

Need to worry about these!

Fall 2011

Example

20% total chip

energy- off

chip

20% total energy –

global on chip

• Need 10-30 fJ/bit for off

chip communication

(current electrical limit is

1000 fJ/bit)

• Need 5-10 fJ/bit for global

on chip communication

(current electrical limit is

100 fJ/bit)

• The global on-chip and

off-chip signaling will

require optics to scale to

ITRS specs

Fall 2011

Motivation: From the website

Communication using light rather

than electrons can consume less

energy for longer links.

• Build optical links

• Transmitter: 10-12 →10-17 J/bit

• Waveguides: low loss, cross-

sections, bends

• Photodetectors: low-

capacitance = high voltage

signal

• Integrate optical links

• Pre-CMOS or post-CMOS?

• Growth or wafer bonding?

Fall 2011

Laser Volume α Energy

How small does the laser have to be?

L

𝐼

𝑞𝑉

1

𝑚3𝑠

𝑁

𝜏

1

𝑚3𝑠

(assume 10 Gbit/sec, InGaAs gain model, Q=100 cavity)

1

𝜏= 𝐴𝑁 + 𝐵𝑁2 + 𝐶𝑁3

Fall 2011

Laser Volume α Energy

10 fJ/bit

Diffraction limit

How small does the laser have to be?

L

𝐼

𝑞𝑉

1

𝑚3𝑠

𝑁

𝜏

1

𝑚3𝑠

(assume 10 Gbit/sec, InGaAs gain model, Q=100 cavity)

1

𝜏= 𝐴𝑁 + 𝐵𝑁2 + 𝐶𝑁3

Fall 2011

Outline

• Motivation

• Previous work

• The nanopatch laser

– Description and operation

– Fabrication challenges

• Conclusion

Fall 2011

Plasmon Lasers

Plasmon

hybridized

nanowire

Plasmonic

crystal defect

Metal-clad

nanofin

CdS/Ag TIR

squares Nanopan

Authors Oulton et al.

(Oct 2009)

Lakhani et al.

(Sept 2011)

Hill et al. (Jun

2009)

Ma et al. (Feb

2011)

Kwon et al.

(Jan 2011)

Mode Volume

(Veff) ??? 2.4 (λ/2n)3 >11 (λ/2n)3 ??? 0.56 (λ/2n)3

Physical

Volume >170 (λ/2n)3 100 (λ/2n)3 >50 (λ/2n)3 ~50 (λ/2n)3 36 (λ/2n)3

Lasing

Conditions

4K, optical

pumping

77K, optical

pumping

10K-RT,

electrical pump

RT, optical

pumping

80 K, optical

pumping

Fall 2011

Metal-optic Lasers Metal-based

VCSEL

Metal-clad

box

Metal-clad

microdisk

Metal-clad

micropost Coaxial Laser Nanopatch

Authors Lu et al. (Jun

2010)

Ding et al.

(Jun 2011)

Nezhad et al.

(Jun 2010)

Hill et al. (Oct

2007)

Khajavikhan

et al.

( Feb 2012)

Yu et al. (Apr

2010)

Mode Volume

(Veff) ??? ??? 4.4 (λ/2n)3 0.38 (λ/2n)3 0.35 (λ/2n)3 0.54 (λ/2n)3

Physical

Volume >4000 (λ/2n)3 >300 (λ/2n)3 110 (λ/2n)3 62 (λ/2n)3 6 (λ/2n)3 6 (λ/2n)3

Lasing

Conditions

RT, electrical

pump, CW

260K,

electrical

pump

RT, optical

pumping

77K, electrical

pump

4.5K, optical

pumping

77K, optical

pumping

Fall 2011

Comparison of Micro/Nano Lasers (Picture Size Normalized to Free-Space Wavelength)

10

2008 2007 2009 2010 2011

Year

Hill

“Gold Finger”

Axel Scherer Group

Microdisk

=650nm

~500nm

Xiang Zhang Group (UCB)

Plasmonic Laser

Kwon, et al

“Nanopan”

~1300nm

Song, et al

Subwavelength Microdisk

Ming Wu Group

(UCB)

Nanopatch

Laser

~870nm

Fall 2011

Outline

• Motivation

• Previous work

• The nanopatch laser

– Description and operation

– Fabrication challenges

• Conclusion

Fall 2011

The nanopatch laser

Gold

Active Gain Region

(InGaAsP)

Gold

Theoretical simulation: Manolatou, C. & Rana, F. ,IEEE J.

Quantum Electron 44, 435–447 (2008).

InP barrier

TiO2

Fall 2011

How it works – modeling

Perfect magnetic conductor

(PMC)

Perfect electric conductor

(PEC)

Device Model

Helmholtz Equation

(from Maxwell’s eqns.) B.C’s Solutions

radial

angular axial

Fall 2011

First two electromagnetic modes

1st mode

Electric Dipole (TM111) mode

gth=2100 cm-1

Q=65

2nd mode

Magnetic Dipole (TE011) mode

gth=1800 cm-1

Q=80

Fall 2011

Gain in semiconductors

𝑄𝑚𝑖𝑛~30

Fall 2011

Smallest fabricated lasers

Vp= 6 (/2n)3

Circular Patch Rectangular Patch

Fall 2011

TM111

TE011

Wavelength Dependence vs. Patch Radius

Electric Dipole (TM111)

Lasin

g

No

Lasin

g Nanopatch radius (nm)

Cav

ity W

avel

engt

h (n

m)

Optical

pumping @

77K

Fall 2011

TM111

TE011

Wavelength Dependence vs. Patch Radius

Electric Dipole (TM111)

Magnetic Dipole (TE011)

Lasin

g

No

Lasin

g Nanopatch radius (nm)

Cav

ity W

avel

engt

h (n

m)

Optical

pumping @

77K

Fall 2011

TM111

TE011

Wavelength Dependence vs. Patch Radius

Electric Dipole (TM111)

Magnetic Dipole (TE011)

Lasin

g

No

Lasin

g Nanopatch radius (nm)

Cav

ity W

avel

engt

h (n

m)

Optical

pumping @

77K

Fall 2011

TM111

TE011

Wavelength Dependence vs. Patch Radius

Electric Dipole (TM111)

Magnetic Dipole (TE011)

Lasin

g

No

Lasin

g Nanopatch radius (nm)

Cav

ity W

avel

engt

h (n

m)

Optical

pumping @

77K

Fall 2011

TM111

TE011

Wavelength Dependence vs. Patch Radius

Electric Dipole (TM111)

Magnetic Dipole (TE011)

Lasin

g

No

Lasin

g Nanopatch radius (nm)

Cav

ity W

avel

engt

h (n

m)

Optical

pumping @

77K

Fall 2011

Radiation Patterns

Fall 2011

Electrically driven nanolasers

gold contact pad

nanopatch laser

air gap

gold via

oxide/semiconductor

gold ground plane

N

P

Fall 2011

Fabricated devices

200 nm

Nanopatch laser diode (PIN junction)

N-contact gold wire

Fall 2011

Laser and LED Characterization optical pumping, 77K, pump=1064nm, 50 ns pulse, 1% duty cycle

1300 1350 1400 1450 15000

5000

10000

15000

A

B A B

Wavelength (nm) Wavelength (nm)

Inte

nsit

y (

a.u

.)

Inte

nsit

y (

a.u

.)

Increasing Electrical Current

Electrical pumping, 300K, CW

PL intensity map

Fall 2011

Outline

• Motivation

• Previous work

• The nanopatch laser

– Description and operation

– Fabrication challenges

• Conclusion

Fall 2011

Simplified Fabrication Flow

Fall 2011

Flip-chip bonding Methods I have used:

• Optically cured epoxy (Norland Optics NOA-81)

• BCB

Other methods available:

• Eutectic (Au/Sn, Au/In) bonding

• Oxide-oxide bonding

Glass slide

epoxy

Epilayer

Si with BCB

III-V

BCB bond recipe:

Spin adhesion promoter and BCB on Si

Spin adhesion promoter on III-V

Flip-chip and bond in N2 oven (need to put

weight on chip and follow baking recipe on

Cyclotene ® website.

Fall 2011

Oxide Deposition

15 nm TiO2

10 nm TiO2

5 nm TiO2

No Gold

• Effect of oxide thickness on PL

• TiO2 has much fewer

defects than Al2O3

• Grown with ALD at 150C

• Prof. Ali Javey’s group has

also studied wet oxidation

growth of oxides for III-V’s

Epilayer

TiO2

Gold

Spectrometer

Fall 2011

Dry etching Options available at Berkeley:

• Ion milling with 500-1000 V Ar+

• RIE etching CH4/H2 (500 VDC)

• (new) ICP Br2 etch (no metal)

Gold

Semiconductor

1000 V Ar+ beam, highly damaging, lots

of surface defects; good profile

RIE etching: 20% CH4 in H2, 500 VDC,

30 mTorr pressure; residue is C-based

polymer, etch damage depth (~60 nm),

absense of Ar makes etch sensitive to

surface contamination, but Ar increases

surface damage

Gold

Semiconductor Polymer

Fall 2011

Wet etching Digital etching technique:

• Oxidize thin shell of semiconductor with peroxide (or equivalent)

• Remove oxide with acid (49% HF works best for us)

• Repeat cycle until desired etch depth (15 nm per cycle)

HSQ

Gold

Epilayer Epilayer

Gold

Fall 2011

Takeaways

• Optical interconnects will inevitably replace copper on

computer chips and enable low-power electronics

• Nanolasers are a promising coherent WDM-capable optical

source with low-power consumption

• The nanopatch laser (world-record setting small size) with

Vp=6 (/2n)3 demonstrates the feasibility and promise of

continuing the development of low-power optical

technology

Fall 2011

Thank You!!!

Questions?

Discussion on the future of nanolasers?

Fall 2011

E-beam Patterning

PMMA/MMA bilayer

(MMA EL9, 4000 rpm 1 min, 150C bake 1 min)

(PMMA A2, 2500 rpm 1 min, 180C bake 1 min)

HSQ

(O2 activation required for adhesion)

(FOX-12 resist 3000 rpm 1 min, 90C back 45 min)

26 X 210 nm fin

Iodine ICP etch

HSQ mask is resistant

Fall 2011

Scaling up optical communication

30-40% CAGR

Key drivers: video, file sharing

Fall 2011

Gain and Quality Factor

Photon generation rate equals photon escape rate

Gain Material Gain Material

threshold gain light velocity confinement factor laser frequency quality factor

Fall 2011

The road to electrical drive

Electrically driven

nanopatch lasers

Problems with

optical cavity

Problems with

electron injection

Introduction of gold

wire

Optical losses due

to doping

Less available gain

Gold wire is too

resistive

Doping is incorrect

No ohmic contact

Epilayer is

inefficient for carrier

recombination

Fall 2011

Outline

• Introduction and Motivation

– Advantages of optical communication

• Laser operation 101

– Basic laser operation

– Challenges of scaling

• The nanopatch laser

– Modeling and building the world’s smallest semiconductor laser

• Beyond lasers

– Ultra-fast nanoLED’s

Fall 2011

Can we get smaller?

• Limitations

• Limited gain from semiconductors

• Metal is very resistive

• Consequence

• Room temperature lasing is hard

• Large threshold needed for population inversion

We may not want to get smaller!

Fall 2011

Stimulated Emission

Semiconductor

Emitter

Fall 2011

Spontaneous Emission

• Spontaneous emission is

slow

– BW ~ 200MHz

– Temporally incoherent

– Spatially incoherent

• Spontaneous emission is

inefficient because the

radiating dipole (x ~ 0.4

nm) is much smaller than

wavelength:

Semiconductor

Emitter

Dipole Length:

x ~ 0.4nm

2x

Fall 2011

Spontaneous Emission

Enhanced by Optical Antenna • By attaching an optical

antenna, spontaneous

emission is enhanced by

• Can be faster than laser at

nanoscale !

• BW of 100s GHz, or even THz

possible

• Temporally incoherent

• Spatially coherent

– Sub-diffraction-limited emitter

Semiconductor

Emitter

2

Spontaneous Hyper Emission

(SHE)

Fall 2011

NanoLED’s Schematic nanoLED Antenna Length: 300 nm

Antenna Width: 50 nm

Fall 2011

Modulation enhancement

0.001 0.01 0.1 1 10 100 1000 1

3

10

30

100

300

1000

3000

Normalized Modal Volume, Vn

Qualit

y F

acto

r, Q

LED

Conventional Lasers

20

40

40

✖

f3dB,opt

Modal Volume, Veff/(λ/2n)3 Vn

Fall 2011

Electromagnetic mode volume

|E|2

x

z y