Nanoscale Explorations of the Insulator-Metal Transition in VO

Nanoscale Explorations

of the Insulator‐Metal Transition in VO2 Jenny Hoffman

Contributors: Yi Yin Changhyun Ko XiangFeng Wang Martin Zech Shriram Ramanathan Gang Wu Tess Willi ams Harvard School of Engineering Xianhui Chen Jeehoon Kim & Applied Sciences USTC Alex Frenzel Harvard Physics Thanks to: Correlated Electron Materials

• High-Tc superconductors: ρ = 0 up to T = 135 Kelvin • Colossal magnetoresistance: B-field changes electrical conductivity by 1000x • Multi-ferroics: magneto-electric-elastic coupling

• Heavy fermion materials: m* = 1000 me • Graphene: m* = 0 • TlililtdiitilTopological insulators: dissipationless sp in; fractional charges; magnetic monopoles

Science 309, 391 (2005) Jannik Meyer

Hoffman Lab JARA, RWTH Aachen Zahid Hasan Physics World Metals: electronic Bloch states are homogeneous, no spatial variation Correlated electron materials: weak screening, s patial inhomo geneit y

Æ We need local probes to study these materials. Hoffman Lab Local Probes

Scanning Tunneling Ultra‐high vacuum Microscope Force Microscope STM

coming on line soon! • spin‐polarized tunneling • atom‐moving

cuprate high‐Tc 122 iron topological metal‐insulator 1111 iron superconductor pnictide insulator transition pnictide Hoffman Lab Local Probes

Scanning Tunneling Ultra‐high vacuum Microscope Force Microscope STM

Themes: • needld loca l s tditudies to un ders tan dblkd bulk proper ties o f corre ltdlated e lec tron ma ter ilials • active manipulation as well as passive imaging • surface probes can be used to investigate bulk properties • for each material studied: • physics: how can we understand it? • technology: how can we apply it? Outline

Strong electron correlations Æ Mott transition MttSMott Syst ems

Vanadium dioxide (VO2) •history • new physics • new applications Superconducting vortices • scanning tunneling microscopy • magnetic force microscopy Outline

Strong electron correlations Æ Mott transition MttSMott Syst ems

Vanadium dioxide (VO2) •history • new physics • new applications Superconducting vortices • scanning tunneling microscopy • magnetic force microscopy Band Theory E Single atom: Crystal: Add momentum information: E

2p Brillouin zone edge: highest unique wavevector k 2s (larger k’s aren’t sampled by the atoms so they are aliased back to smaller k’s)

λmin kmax 1s Band Theory: Metal E Single atom: Crystal: Add momentum information: E

2p Brillouin zone edge: highest unique wavevector k 2s (larger k’s aren’t sampled by the atoms so they are aliased back to smaller k’s)

λmin kmax 1s Band Theory: Insulator E Single atom: Crystal: Add momentum information: E

2p Brillouin zone edge: highest unique wavevector k 2s (larger k’s aren’t sampled by the atoms so they are aliased back to smaller k’s)

λmin kmax 1s Band Theory: Metal E Single atom: Crystal: Add momentum information: E

2p Brillouin zone edge: highest unique wavevector k 2s (larger k’s aren’t sampled by the atoms so they are aliased back to smaller k’s)

λmin kmax 1s Band Insulator: Peierls Transition E Molecule: Crystal: Add momentum information: E

2p Brillouin zone edge: highest unique wavevector k 2s (larger k’s aren’t sampled by the atoms so they are aliased back to smaller k’s)

new λmin doubles kmax halves 1s Band Theory: Metal E Single atom: Crystal: Add momentum information: E

1s Add e‐e Correlations E Single atom: Crystal: Add momentum information: E

e e e e e e e e

1s Add e‐e Correlations E Single atom: Crystal: Add momentum information: E

e e e e e e e e

1s Add e‐e Correlations E Single atom: Crystal: Add momentum information: E

e e e e e e e e

1s Mott Insulator E Single atom: Crystal: Add momentum information: E

e e e e e e e e

1s Mott Transition

localized

delocalized

further delocalized Outline

Strong electron correlations Æ Mott transition MttSMott Syst ems

Vanadium dioxide (VO2) •history • new physics • new applications Superconducting vortices • scanning tunneling microscopy • magnetic force microscopy Mott Insulator: NiO

N. F. Mott, Proceedings of the Physical Society of London A62, 416 (1949) Mott Insulator: Hydrogen

best data metal? to date Mott transition? insulator

Deemyad & Silvera, PRL 100, 155701 (2008) Mott Transition: Cold Atoms

V0 = potential depth of lattice

tighter lattice,

increasing V0 sharper interference

loss of coherence = Mott transition Greiner, Nature 415, 39 (2002) Mott Insulator: Undoped Cuprate

max Bi2Sr2CaCu2O8+d (Tc ~ 93 K) BiO Each Cu is singly occupied Æ half-filled band, should be metal? SrO Strong Coulomb repulsion CuO2 Æ Mott insulator Ca

CuO2 SrO BiO

BiO SrO

CuO2 Ca

CuO2 SrO BiO Mott Transition: Dope the Cuprate

max Bi2Sr2CaCu2O8+d (Tc ~ 93 K) BiO SrO

CuO2 Ca

CuO2 SrO BiO

BiO SrO

CuO2 Ca

CuO2 SrO BiO Mott Transition: Dope the Cuprate

max Bi2Sr2CaCu2O8+d (Tc ~ 93 K) BiO SrO

CuO2 Ca

CuO2 SrO BiO

BiO SrO

CuO2 Ca

CuO2 SrO BiO Cuprate Phase Diagram s 3-dim cuprate phase diagram “strange metal” ) Boyer, Nat Phys Nat Phys (2007) Boyer, 1.2 density of state mm

energy (V)(meV) c

Ω 0.8

0.4 sistivity (m sistivity

ee 0 r 0 200 400 600 800 T (K)

T antiferromagnetic states insulator ff

density o energy (meV)

d-wave superconductor B doped Mott insulator x (doping) becomes metal Cuprate Phase Diagram

3-dim cuprate phase diagram “strange metal”

“Here be dragons”

T antiferromagnetic insulator

d-wave superconductor B doped Mott insulator x (doping) becomes metal Cuprate Phase Diagram

3-dim cuprate phase diagram “strange metal”

“Here be dragons”

T antiferromagnetic insulator

d-wave superconductor B doped Mott insulator x (doping) Interest in High‐Tc Cuprates

funding agency frustration sets in? (price of helium skyrockets)

J Supercond Nov Magn 21, 113 (2008) Outline

Strong electron correlations Æ Mott transition MttSMott Syst ems

Vanadium dioxide (VO2) •history • new physics • new applications Superconducting vortices • scanning tunneling microscopy • magnetic force microscopy Outline

Strong electron correlations Æ Mott transition MttSMott Syst ems

Vanadium dioxide (VO2) •history • new physics • new applications Superconducting vortices • scanning tunneling microscopy • magnetic force microscopy 1959: VO2 Metal‐Insulator Transition

T (Kelvin) 500 250 167 125 100 83

Morin, PRL 3, 34 (1959) VO2 Metal‐Insulator Transition T (K) Tetragonal Metal 105 Heating V4+ Cooling 104 ) ΩΩ 3

R ( 10 340 K

Monoclinic Insulator 102

300 320 340 360 380 T (Kelvin)

Also transitions as a function of • stress (~38 kbar) • opti cal ex ci tati on • applied voltage or current (~107 V/m ??) Æ up to 5 orders of magnitude change in conductivity VO2 Metal‐Insulator Transition: Physics T (K) Tetragonal Metal

3dπ (V-O-V)

3d|| (V-V)

340 K Band insulator? E Monoclinic Insulator ? k

3dπ (V-O-V) • Goodenough, Phys. Rev. 117, 1442 (1960) • Wentzcovitch, PRL 72, 3389 (1994) ? E 07eV0.7 eV F Mott insulator?

3d|| (V-V) U

e e e e

• Zylbersztejn & Mott, PRB 11, 4383 (1975) Goodenough, J.Solid State Chem 3, 490 (1971) Outline

Strong electron correlations Æ Mott transition Mott Systems

Vanadium dioxide (VO2) •history • new ppyhysics • new applications Superconducting vortices • scanning tunneling microscopy • magnetic force microscopy VO2 Recent Revival Mott vs. Peierls (electronic vs. structural)? (Mott is more interesting) • m* increases, approaching metal Æ insulator transition Qazilbash, Science 318, 1750 (2007) • Δt = 80 fs, li mit ed b y exactl y ½ per io d o f s truc tura l p honon Cavalleri, PRB 70, 161102 (2004) Large Gap at Low T

4000 cm-1 ~ 0.5 eV

Redistribution persists up to 6 eV ~ 7 0,000 K

Gap = 0.6 eV ~ 7,000 Kelvin

Qazilbash, PRB 77, 115121 (2008) Phillips, RMP (2010); arXiv:1001.5270 m* Enhancement Approaching Insulator

less more conducting conducting Qazilbash, Science 318, 1750 (2007) Insulator ÆMetal Transition Limited by Phonon

response time >> excitation time response time ~ excitation time

excitation response

½ phonon period

Æ importance of band nature of transition Cavalleri, PRB 70, 161102 (2004) VO2 Recent Revival Mott vs. Peierls (electronic vs. structural)? (Mott is more interesting) • m* increases, approaching metal Æ insulator transition Qazilbash, Science 318, 1750 (2007) • Δt = 80 fs, li mit ed b y exactl y ½ per io d o f s truc tura l p honon Cavalleri, PRB 70, 161102 (2004)

Cr/Au

metallic state

transition occurs at I < 3 mA

H-T Kim, APL 86, 242101 (2005)

Also: Qazilbash et al infer decouping from nanoscale IR and XRD (but not simultaneous expts)… APS March Meeting (2010) VO2 Recent Revival Mott vs. Peierls (electronic vs. structural)? (Mott is more interesting) • m* increases, approaching metal Æ insulator transition Qazilbash, Science 318, 1750 (2007) • Δt = 80 fs, li mit ed b y exactl y ½ per io d o f s truc tura l p honon Cavalleri, PRB 70, 161102 (2004)

Can it be entirely E-field driven or need heat too ? (would like zero quiescent current) • 3-terminal (field-effect) devices show transition (but questions of breakdown voltage…) Stefanovich, JPCM 12, 8837 (2000) Qazilbash, APL 92, 241906 (2008) • thermal modeling says Joule heating is not enough to drive transition Gopalakrishnan, J . Mat . Sci . 44 , 5345 (2009) Transition Triggered by E‐field ?

1) Source-drain voltage: observe switching in 1. 5 ns. Compute that time for Joule heating to raise T > 340K would be 900 ns. Æ Conclude that E-field, not Joule heating drives transition

2) Gate voltage: observe polarity asymmetry Æ Conclude that electron injection (but not hole injection) drives transition

Caveats: 1) Thermal modeling relies sensitively on nanoscale details. 2) Possible heating from electrical breakdown?

Stefanivich, JPCM 12, 8837 (2000) Transition Triggered by E‐field ?

IR absorption measured from above

integgprated absorption = increasing metallicity

No source-drain current Æ conclude no Joule heating Polarity effect Æ conclude true gating of insulator-to-metal transition Caveat: sample doesn’t recover after removal of gate voltage

Æ barrier breakdown effect, rather than VO2 effect?

Qazilbash, APL 92, 241906 (2008) Hoffman Lab Force Microscope

• 2 K to above 340 K • 5 T vertical field • lateral coarse motion (3 mm x 3 mm) allows imaging of isolated features in addition to bulk materials • high-resolution, easily modelable tips fabricated in house via focused ion beam

Radius of curvature : 15-25 nm Cone half-angles : 1-3° 12 cm Aspect ratios : 12-18

Versatility: • vertical or lateral force measurement • magnetic tips for magnetic imaging and manipulation • conducting tips for local conductivity imaging and switching • vertical cantilevers for friction imaging built by Dr. Jeehoon Kim Hoffman Lab Force Microscope

50-pin feedthrough Turbo pump

Angle valve

Al hexagon plate

Wooden table filled with 2000-Lb lead bricks

cryostat Air spring Eddy current Damping system

pit MFM head location Bulk vs. Local Local geometry: Bulk thermal transition 5 cantilever: k ~ 40 N/m 10 c Heating Cr/Au n‐type Si Cooling coated tip 104 ) Ω

2 3 RR ~10

R ( 10 Pd VO 2 VO2 2 SiOx 10 n-type Si n‐type Si

300 320 340 360 380 V T (Kelvin) Single Point IV (conducting AFM): Look for bulk voltage transition 2 10 298K 293 K 101 note ) multiple AA 0 10 jumps

I (m -1 10 2 100 x 100 μm 2 10-2 500 x 500 μm 01234 V (V) Force Dependence

293 K (uA) tt Curren

•large minimum force for reliable contact

•> Fmin, little dependence on F Training & Repeatability

293 K

Training voltage: Typically several high voltage sweeps necessary before IV hysteresis loops roughly stabilize with ~5% jitter

Origin of training? 1. tip 2. junk on surface

3. SiOx interface layer 4. O diffusion / annealing of grain boundaries Resistivity Ratio

Bulk thermal transition Local voltage transition

105 Heating 293 K Cooling 104 ) ΩΩ

2 3 RR ~10 RR ~10

R ( 10 Pd

VO2 2 10 n-tSitype Si

300 320 340 360 380 T (Kelvin) Æ R in “metal” state dominated by series R ~13 kΩ

Origin of series R ? 1. cantilever ( < 400 Ω ) 2. tip-samplttle contact

3. single grain of VO2 ( 19 Ω ) 4. grain boundary R ( ~1 kΩ ) 5. Si substrate ( 160 Ω )

6. SiOx layer ( < 50 Ω ) Imaging the Metal‐Insulator Transition

Topography

40 nm

0 500 nm Imaging the Metal‐Insulator Transition

Current map Tip bias: 3.45 V

171 μA

0 500 nm Imaging the Metal‐Insulator Transition

Current map Tip bias: 3.95 V

171 μA

0 500 nm

Æ seed grain is the one with highest conductivity in the insulating state Imaging the Metal‐Insulator Transition

Current map Tip bias: 4.44 V

171 μA

0 500 nm Imaging the Metal‐Insulator Transition

Current map Tip bias: 4.93 V

171 μA

0 500 nm Imaging the Metal‐Insulator Transition

Current map Tip bias: 5.43 V

171 μA

0 500 nm Imaging the Metal‐Insulator Transition

Current map Tip bias: 4.93 V

171 μA

0 500 nm Imaging the Metal‐Insulator Transition

Current map Tip bias: 4.44 V

171 μA

0 500 nm Imaging the Metal‐Insulator Transition

Current map Tip bias: 3.95 V

171 μA

0 500 nm Imaging the Metal‐Insulator Transition

Current map Tip bias: 3.45 V

171 μA

0 500 nm VO2 Questions Mott vs. Peierls (electronic vs. structural)? (Mott is more interesting) • m* increases, approaching metal Æ insulator transition Qazilbash, Science 318, 1750 (2007) • Δt = 80 fs, limited by exactly ½ period of structural phonon Cavalleri, PRB 70, 161102 (2004)

Can Mott exist w/o Peierls? (Mott would be even faster, wouldn’t crack crystal) • 46 years (1959 – 2005): answer is NO • Raman spectroscopy Æ yes? H. T. Kim, APL 86 , 242101 (2005) • Local IR microscopy (electrical) & x-ray diffraction (structural) Æ yes? Qazilbash, unpublished APS March talk (2010) Æ AFM detects structural; CAFM detect electronic; simultaneous independent info Simultaneous I and Z detection Simultaneous I and Z detection VO2 Questions Mott vs. Peierls (electronic vs. structural)? (Mott is more interesting) • m* increases, approaching metal Æ insulator transition Qazilbash, Science 318, 1750 (2007) • Δt = 80 fs, limited by exactly ½ period of structural phonon Cavalleri, PRB 70, 161102 (2004)

Can it be entirely E-field driven or need heat too ? (would like zero quiescent current) • 3-til(fildterminal (field-effect) dev ices s how trans ition (bu t ques tions o f brea kdown vo ltage… ) Stefanovich, JPCM 12, 8837 (2000) Qazilbash, APL 92, 241906 (2008) • thermal modeling says Joule heating is not enough to drive transition Gopalakrishnan, J. Mat. Sci. 44, 5345 (2009) Æ scanning gate microscopy Scanning Gate Microscopy

Topinka, Nature 410, 183 (2001) Outline

Strong electron correlations Æ Mott transition MttSMott Syst ems

Vanadium dioxide (VO2) •history • new physics • new applications Superconducting vortices • scanning tunneling microscopy • magnetic force microscopy VO2 micro‐bolometers for IR imaging

IR impinges here

SiN + VO2

50 μm 0.5 μm y wire 2.5 μm

http://www.security-int.com

x wire monolithic tittransistor

http://www.thermodelta.hu

• pixels size as small as ~ 25 μm • pixel arrays up to 640 x 480 VO2 windows?

Thin film VO2 for energy-saving “thermochromic windows”. When VO2 is colder than 68° C (154° F) it is transparent. When it is heated a few degrees higher, however, it becomes reflective. SthbiiditSo the basic idea is to crea te w idindows thtthat are transparen ttlt at lower temperatures and then block out sunlight when the temperature rises.

Patent # 4,401,690 Charles Greenberg Submitted on 02/01/1982 Issued on August 30, 1983 VO2 Fire Detectors ??

Hampton Inn 3:30am (last night)

4:00am Just a faulty smoke detector

Could a VO2 detector have prevented this?

Verkeles, Sensors & Actuators A 68, 338 (1998) VO2 tunable metamaterials

“Invisibility cloak”: fabricate copper patterns on standard circuit board with fixed dielectric

Problem: narrow, fixed bandwidth

Proposal: use VO2 substrate Æ tunable dielectric Æ tblttunable metama tilbdidthterial bandwidth David Smith, Duke University Prototype:

Driscoll, APL 93, 024101 (2008) Memristors V

Resistor Capacitor dV = RdI = Vdt dq = CdV Applications φ dd I dq = Idt q • Memory circuits

• Digital logic circuits IdInduct or MitMemristor (e.g. FPGA = Field Programmable dφ = LdI dφ = Mdq φ Gate Array) L. Chua, IEEE Trans. Circuit Theory 18, 507 (1971) • Analog circuits D. Strukov, Nature 453, 80 (2008) (e.g. neural learning networks) Integral form ∫∫dMqdqϕ = () Differential form dddtVϕ ϕ / Mq()== = dq dq/ dt I A memristor remembers its resistance, based on the history of charge flowing through it. VO2 Memristors

Mechanism 1: TiO2 (insulating) OiitO vacancies migrate 10 nm oxygen migration when voltage is applied TiO2-x (conducting)

Mechanism 2: Ti4O7 conductive channel formation apply V

TiO2 Æ messy, disordered… problems with reproducibility 10 nm Kwon,,,() Nat Nano 5, 148 (2010)

Mechanism 3: phhhase change

Æ potentially cleaner, reproducible nt (mA) Curre

5mm Volts Driscoll, APL 95, 043503 (2009) Memristive Writing ? Step 1: “write” with 13 V tip bias

Cr/Au n‐type Si coated tip 293 K

VO 2 SiO n‐type Si x V

Step 2: read current (local resistivity) with 3 V tip bias

13 μA

500 nm 0 Conclusions Vanadium Oxide: our best Mott ppyglayground • controlled local transition (single grain, 10s of nanometers)

500 nm • after training, hysteresis is repeatable

• can “it”“write” a new res itittistive state

500 nm To Do Next • Simultaneous imaging of structural & electronic state Æ prove decoupling?

• Scanning gate imaging of VO2 Æ prove E-field trigger ? • “Write” the Mott transition in vanadium oxide for memristive applications

Outlook • correlated electron systems Æ local inhomogeneity Æ need local probes • surface probes can access bulk properties, e.g. vortex pinning • passive imaging Æ active manipulation