Coherence in Superconducting Materials for

Quantum ComputingDavid P. Pappas

Jeffrey S. Kline, Fabio da Silva, David WisbeyNational Institute of Standards & Technology,

Electronics & Electrical Engineering Laboratory, Boulder, CO

Collaborators

Will Oliver, Paul Welander – MIT/LL

Ray Simmonds, Kat Cicak, Josh Strong NIST, Boulder

Matthias Steffen, IBM Watson

Kevin Osborn, LPS, MD

John Martinis, Haohua Wang, UCSB

Rob McDermott, U of W

Sponsors

The quantum computing challenge

QubitPrepare

Measure

QubitPrepare

Measure

QubitPrepare

Measure

Implementations

PhotonsIon trapsNeutral atomsNMR~~~~~~~~~~~~~~

~~~~~~~~~~~~~~Spins in semiconductorsQuantum dots~~~~~~~~~~~~~~

Superconducting:~~~~~~~~~~~~~~~ChargeFluxPhase

Isolation Coupling

Decoherence:

external – radiation, heat, acoustic…internal – materials, crosstalk…

interact

interact

Superconducting qubit measurement setup

• Ante

• Dilution Refrigerator

• Low temperature, < 50 mK

• RF measurement

• Low power ~ 1 photon of energy

in cavity

• Improves coherence

• Removes quasiparticles in

superconductor

• Reduces thermal radiation

• Hurts coherence:

• Low-energy, two-level

excitations in amorphous

materials

The Josephson Junction

0

ie0

1.7 nm

• Building block of superconducting quantum bits (qubit)• Josephson relations (’62, ‘73)

Al

Al

amorphous AlOX

• Not ohmic = > I periodic in d• Voltage only when phase is changing

• System is nonlinear for high I

e2

V

)sin(II 0

IU cos

TEM photo

Types of qubits

“Charge”

“Flux”

“Phase”

You & Nori, Physics Today, November (2005)

Logic

Non-linear oscillatorExcited |1>

vsGround state |0>

IslandCharged

vs.Not charged

Current circulationLeft vs

Right

Anatomy of a conventional superconducting circuitMaterials perspective

Tunnel barrier

Wiring

Insulator

Substrate

Material Preparation method

Tunnel Barrier AlOX Thermal

Wiring Nb or Al Sputtered

Insulator SiOX CVD

Substrate Si/SiOX Thermal

Traditional

Conventional materials are usedfor a lot of really good reasons…

• Si substrate with thermal amorphous a-SiOX on top

– Smooth, standard lithography, inexpensive

• a-SiOX insulators – CVD

– Smooth (no pinholes), low T, easy

• a-AlOX tunnel barrier – thermal or plasma oxidation

– Smooth, no pinholes, low T, easy, self-limiting

• Nb or Al wiring – sputter deposit, polycrystalline

– Low temperature, smooth, relatively high TC

Need strong motivations for change …

8

Short lifetimes of quantum information in solid state superconducting qubits

• Relatively short lifetimes and operation cycles

• Need lifetime/gate operation time > 1000

0.5

0.25

00 100 200

T1 = 23 ns

Pro

b. |1

> s

tate

Meas. delay (ns)

0

1

Lifetime

“Rabi” oscillation

Outline

• Electrical model of a phase qubit• Two Level Systems (TLS) as loss mechanism

– substrate & insulators a-SiOX

– tunnel barrier a-AlOX

• Test structures for materials analysis• New directions in materials

– Improved substrates a-Si & removal– Crystalline barriers Al2O3

• Recent progress

LCR electrical model for phase qubit

=

CJ~1-100 x10-12LJ~sinf

JJ CL1

0

Inte

nsi

ty

JCVG )(T :state 1 of Lifetime 1

G(V)

• Quality factor – Energy stored/Energy lost/cycle

• Q = = w0/Dw

T1 = Q/w0

• Delectric loss tangent:

• tand = Im(e)/Re(e)

= 1/Q

Rjunction – non-linear QP tunneling - ?

Rdielectric – bound dipole relaxation ~ ?

Junction & insulators

What can we easily measure & optimize?

frequency

8GHz @ s 125~T ) tan( ,10 ~ Q :Goal 16

Loss in amorphous materials (SiOX-OH-)

• Low energy displacements of dipoles, saturate at high T, P• Lose energy through phonon creation

– tan = d 3x10-3, Q ~333, T1~40 ns

• Approaches: 1) Reduce or eliminate dielectrics

2) Optimize mtls. – e.g SiN, a-Si…

Schickfuss & Hunklinger, (1974)

E d

++++ ++++

_ _ _ _ _ _ _ _

Minimize & optimize dielectric - qubit Rabi oscillations

Rabi oscillations > 600 ns !!

Sapphire substrate + SiN insulator:

Kevin Osborn Group

SiNX pillar from high-stress film

Al film

SiNX

200 nm

Optimized SiNx for coherent quantum circuits

Qi=25,000

Qi=1,400

Loss Tangent for SiNx films

The loss tangent is sensitive to PECVD growth!

smooth etch profile from HDP CVD filmprecursor ratio: N2/SiH4 = 1.8x-ray reveals polycrystalline orderstress = 600 MPa compressiveTgrowth = 300 C

other labs:NIST,UCSB

Optimize dielectrics with simple L-C circuits

L

C

LC – parallel plate C CPW

Material Q=1/tand

Si(111) 200,000

Sapphire – Al2O3 160,000

a-Si:H 45,000

a-SiN 10,000

a-SiOX 3,300

sT HSia 2.1:1

O’Connell, APL (2008)

Predicts:

Substrate

insulator

Other approach – remove dielectricsSimmonds, Strong, Cicak et. al, NIST Boulder (2008)

• Vacuum gap capacitor with an inductor

Q

Before dielectric removedSiN, 100 nm x 8000mm2

600

After dielectric is removed 40,000

=> Flexible circuit - allows us to test the loss in a junction under identical conditions

Add a 1.5 nm, 10 mm2 a-AlOX JJ to the circuit

Q

Before dielectric removedSiN 100 nm x 8000mm2

600

After dielectric is removed 40,000

With Josephson Junctiona-AlOX, 1.5 nm x 10mm2

400

• Generally understand dielectric problem – Improve & Reduce• Significant loss in the amorphous AlOX junction

• 1.5 nm thick – very strong coupling• Focus on tunnel barriers

Tunnel barrier material characterizationQubit spectroscopy

• Increase the bias voltage (tilt)• Frequency of |0> => |1> transition goes down

Splittings

IncreaseI bias

Splittings in charge qubit - Cooper-Pair Box

Vg

Cg (Ec,EJ)

1 mm

B

Al/AlOx/Alisland

BVg

gate

islandgate

junction

junction

Z Kim et al., Physical Review B 78, 144506 (2008).

Effects of splittings

• Quench Rabi Oscilations – strong coupling to qubit• Reduces the measurement fidelity

Rabi oscillations

Spectroscopy

Origin of spectroscopy splittings

• Individual, strongly coupled TLS’s in barrier

• Distribution of excitation energies - amorphous AlOX

Density of splittings ~ 1/GHz/mm2 in 1.5 nm thick junction

(1) Reduce materials where possible

(2) Improve materials by eliminating TLS’s

13 um2 junction

• Fewer splittings, large gaps

• stronger coupling

70 um2 junction

• More splittings, small gaps

• weak coupling

1) Reduce materials where possibleSteffen, et. al PRL (2006)

• Reduce size of junctions in qubits

• increases f0 due to smaller capacitance

• Add high quality external capacitor to bring f0 down (SiN, a-Si)

• T1 ~ 170 ns (SiN) & 600 ns (a-Si:H)

• Factor of 2 shorter than expected - Still have a-AlOX in barrier

Growth of single-crystal Al2O3 (sapphire) tunnel barrier

4×10-6 Torr O2, Al 10-6 Torr O2

Epitaxial Re/Al2O3

Re

@ 850 CAl

Amorphous AlOX

@ RT

Epitaxial Al2O3

@ 800 C

Polycrystalline Al

@ RT

• Rhenium bottom electrode:• Superconducting – TC ~1 K• hcp - lattice match Al2O3• high melting T

(2) Improve JJ’s with crystalline barriers - Al2O3 & MgO

• Good - High sub-gap resistance

• First high quality junctions

made with epitaxial barrier

• Fabricate into qubit

Re(0001)

Al2O3

Al

Re

Al

I-V curve

20 mK

V(mV)

• T1 > 500 ns

– best for SiO2 insulator & large junction

– No external capacitance• Splitting density reduced

– ~3-5 times lower than amorphous barrier of same area

Qubit with 25 mm2 epitaxial Al2O3 junctionKline, et. al, Supercond. Sci. Tech. 22, 015004 (2008)

Summary & OutlookMaterials in superconducting qubits

12 Qubit Test Die Layout

Bias coil Qubit loop

DC-SQUID

Two level systems in junction

Amorphous AlO tunnel barrier

• Continuum of

metastable vacancies

• Changes on thermal cycling

• Resonators must be 2 level,

coherent with qubit!

I

What we need:

Crystalline barriera-Al2O3

Poly - Al

Poly- Al

Existing technology:

Amorphous tunnel barrier a –AlOx – OH-

No spurious resonatorsStable barrier

Amorphous Aluminum oxide barrierSpurious resonators in junctionsFluctuations in barrier

Silicon

amorphous SiO2

Low loss substrate

Design of tunnel junctions

SC bottom electrode

Top electrode

Q: Can we prepare crystalline Al2O3 on Al?

Binding energy of Al AES peak in oxide60

59

58

57

56

55

54

900800700600500400300Annealing Temp (K)

AE

S E

nerg

y of

Rea

cted

Al (

eV)

Al in sapphire Al203

Metallic aluminum

Aluminum Melts

68

10 Å AlOx on Al (300 K + anneal) 10 Å AlOx on Al (exposed at elevated temp.)

Anneal the natural oxides Oxidize at elevated temp.

A: No – need high temperature bottom wiring layer

Motivations – New wiring materials• Conventional Al, Nb:

– Surface oxides with spin polarized traps• 1/f flux noise, dephasing times, density ~ 1017/m2

• Alternative materials:– Re: resists oxidation, high melting T, hcp lattice => Al2O3,

– Al passivated with Re or Ru => resists oxidation

Koch, Clark, di Vincenzo(PRL 2007)

e- traps Kondo traps

Faoro, Ioffe PRB (2007)

Coupled TLS

McDermott, et. al (2007)

Improvement of junctionsseen in spectroscopy of 01 transition

T = 25 mK

Amorphous barrier70 m2

Epitaxial barrier70 m2

• Density of coherent splittings reduced by ~5

in epitaxial barrier qubits

Source of Residual TLFs: Al-Al2O3 interface?

Electron Energy Loss Spectroscopy (EELS) from TEM shows1. Sharp interface between Al2O3 and Re2. Noticeable oxygen diffusion into Al from Al2O3

1. Indicates presence of a-AlOx at interface2. Will “heal” pinholes

Distance (μm)

Oxy

gen

cont

ent

Al2O3White is oxygen

Need to improve top barrier interface!

• Interfacial effect• ~1 in 5 oxygens at Al interface• Agrees with reduced splitting density

~1.5 nm

epi-Re interface

non-epi Al interfaceOxygen

Re

Al

a-AlOx

0

5

10

15

20

25

0 100 200 300 400 500

Al/a-AlOx/Al

V (uV)

Al/a-AlO/Al

0

4

8

12

0 200 400 600

Re/c-Al2O3/Al

V (uV)

Re/c-AlO/Al

0

5

10

15

20

0 200 400 600

Re/c-MgO/Al

V (uV)

Re/c-MgO/Al

a: Amorphousc: Crystalline

Supports conclusion that Al top electrode “heals” pinholes

substrate

Al top electrodeTunnel barrierBottom electrode

Top electrode mattersAl top electrode always gives good I/V

0

10

20

30

0 200 400 600

Re/c-Al2O3/Re/Al

V (uV)

Re/c-AlO/Re

substrate

Re top electrodeTunnel barrierBottom electrode

=> Pinholes in tunnel barrier

Re on top makes JJ leaky

Electrical Testing Summary & ComparisonPhase qubits

Materials Wiring & barrier

Insulator T1

(ns)

T2*

(ns)

Splitting density(N/GHz/mm2)

Reference

Al/AlOx/Al

1 mm2 w/shunting C

min-SiNx 110 90(160)

1 Steffen - tomographyPRL 97 050502

Al/AlOx/Al

13 mm2

min-SiNx 500 150 1 Martinis Dielectric lossPRL 95 210503

Al/AlOx/Al min-SiO2 170 * 1 Simmonds 2005

Re/Al2O3/Al epi-junction max-SiO2 150 90 0.2 PRB 74 100502

12 qubit - Re/Al2O3/Al

49 mm2

max-SiO2 200-400 * 0.2 Submitted APS08

12 qubit - Re/Al2O3/Al

49mm2

min-SiO2 500 140 0.2 Submitted APS08

12 qubit – Re/MgO/Al 80 50 0.4 New results

Goals1. Inter-laboratory compatibility

– Infrastructure - 6”-wafer chamber for epitaxial trilayers

• Develop 6” substrate capability

• Re/Al2O3/Al, Re/Al2O3/Re

– Supply samples to flux qubit, 6” wafer fabrication facililty

2. Extend work on epitaxial tunnel barriers to flux qubits

– Continue on barriers at chip level

• Chip level

– Develop JJ and qubit circuits compatible w/flux qubits

– study fully epitaxial systems

3. Study new materials for wiring layers

– Al/Ru capping with anneal

– Push to understand flux noise and wiring surfaces

• “Medium” K dielectrics?• Si• SiN• Al2O3

• MgO• Diamond• ZrSiO• CaO• SiC

Þ Need to use thicker insulators

• “low” K dielectrics? • doped SiOx (F, C• Porous SiOx• Spin-on polymers (HSQ)

Probably not

new

Other potential new insulators – from VLSI world?

New directions

tunnel barrier

insulator

wiring

substrate

Substrate Sapphire(Al2O3)

Crystalline Expensive, difficult to work with, can be atomically rough

Wiring Re, Al/Ru Annealed Complicated, hard to prepare, Hi-T

Insulator SiN, a-Si, Al2O3

SputteredEpitaxial

High T, adhesion, processingHomogeneity, rough

Barrier Al2O3 Epitaxial High T, homogeneity, rough

Materials Difficulties

CMOS

• TLS bath saturates at high E (power), decreasing loss

Schickfus and Hunklinger, 1975

Two-level systems in a-SiO2

E d

SiO2 - Bridge bond

UDAmorphous material has all barrier heights present

High E

Low ELow E

2 RSiO2Ceff 27ns

~T

RSiO2

=2.1kW

Temperature Dependence of QQ also decreases at low temperature!

Problem - amorphous SiO2

Why short T1’s in phase Josephson qubits?

Dissipation: Idea - Nature:At low temperatures (& low powers)environment “freezes out”:

dissipation lowers

dissipation increases, by 10 – 1000!

Change the qubit design:

Þ find better substrates

Þ find better dielectric & minimize insulators in design

Common insulator/substrate materials

• SiOX

– Bridge bond, unstable• Amorphous films have uncompensated O- , H, OH-

• Si3N4

– N has three bonds – more stable• Amorphous films, still have uncompensated charges, H• 20% H for low T films, ~ 2% H in high T films

• Al2O3

– Amorphous – high loss, similar to a-SiO2, has H, OH- in film– Single crystal (sapphire) - Very low loss system

Insert qubit pic here

Qubit LStripline (C-SiOX )

Josephson Junction(L&C)

=> Measure “Q” of simple LC resonators

Qubit has SiO2 Cap in || with J.J. & around lines

SiOX AlOx

Superconductor - Aluminum

I

Tunnel junction a- AlOx-OH-

Found improvements due to optimized materials in insulators

Tunnel barrier materials