F. Kreupl et al. 1

Carbon-based Resistive Memory

Franz Kreupl, Rainer Bruchhaus+, Petra Majewski , Jan B. Philipp,

Ralf Symanczyk, Thomas Happ*, Christian Arndt*, Mirko Vogt*,

Roy Zimmermann*, Axel Buerke*, Andrew P. Graham*, Michael Kund

Qimonda AG,

+Qimonda North America, *Qimonda Dresden GmbH & Co. OHG franz.kreupl@qimonda.com

F. Kreupl et al. 2

Outline of Presentation • Introduction and Motivation • Basic Switching in Carbon • Carbon Allotropes • Carbon Nanotubes • Graphene-like Conductive Carbon • Insulating Carbon • Conclusions

F. Kreupl et al. 3

Introduction and Motivation • Resistive Memories like

– Phase Change (GexSbyTez,.... ) – Nano-Ionic, (CBRAM-like, e.g. Ag2S, Ag-GeSe, Cu2S....) – Transition-Metal-Oxide ( NiO, TiO2,.....) are very promising memory technologies

• but are binary, ternary or even more complex materials: – what about scalability – how do they respond to volume/surface ratio – what about variability, if scaled

• Are there other, simpler materials available?

F. Kreupl et al. 4

Introduction and Motivation • The available current in a memory cell

is given by the select device (FET, diode) An upper limit may be estimated by:

j= e*Nd*vsat= e*1019*107 = 16 MA/cm2

• Typical currents in phase change

memories are ~10 MA/cm2 Are there options or materials

which enable switching at low currents (some µA)?

F. Kreupl et al. 5

Introduction: Carbon Memory sp3 sp2

I2 > I1

high conductance low conductance

I1

sp2 to sp3 conversion of disordered graphitic carbon (phase change of carbon) inherently scalable to atomic scale (no phases of different materials)

F. Kreupl et al. 6

Basic Switching in Carbon I1

TEM image by courtesy of J. Huang et al., Nano Letters 2006 Vol. 6, No. 8 pp. 1699-1705

• Current changes structure and resistance • Resistance changes by a factor of ~ 100 • High Low well known from e-fuses • New: switch to disorder by short pulse

I + -

F. Kreupl et al. 7

Short Laser Pulses on Graphite

Bonelli et al., Laser-irradiation-induced structural changes on graphite, Phys. Rev. B 59, 13513 (1999)

• Short laser pulse induces disorder (D-band) • D-band overlaps with sp3-peak at 1332 cm-1 • Diamond cubic phase observed by

e-beam diffraction Disordered, quenched state by short energy pulse

0 1000 2000 30000

5001000150020002500

Inte

nsity

(a.u

)

Raman shift cm-1 0 1000 2000 30000

500

1000

1500

Inte

nsity

(a.u

.)

Raman shift cm-1

pulse 20 ns

1580 cm-1 G=1580 D=1355

before after overtones

F. Kreupl et al. 8

3 ps Laser Pulse on Graphite

source: Anming Hu, PhD Thesis, U Waterloo, 2008

sp3 sp

632 nm 325 nm

sp3

• The shorter the laser pulse the more disorder Disordered, quenched state by short energy pulse

F. Kreupl et al. 9

Short Energy Pulses

• Fluence of the laser pulse at graphite ablation threshold =~ 285 mJ/cm2

Energy density (energy/volume) = 95 KJ/cm3 • Energy density for a wire subjected to a current:

E/V = j2ρt = 1GA/cm2 ·1mΩcm · 20 ns = 2 MJ/cm3

Disordered, quenched state by short current pulse

K. Sokolowski-Tinten et al., CLEO 2000

F. Kreupl et al. 10

Short Current Pulse

temperature distribution in carbon filament after 1 ns current pulse with 1.7 GA/cm2 : Tpeak ~ 3900 K; rapid cool down (0.05 ns)

carbon cell

F. Kreupl et al. 11

Carbon Nanotubes

• Nanotubes have a length dependent switching current • 25 nm long tubes need ~ 100µA @ 1.5 V

and have no phonon-limited transport • tubes > 200 nm need ~ 30 µA @ 4V (phonon-limited) Select device needs to handle ~ 30 µA and 4-8 Volt

0 1 2 3 40

20µ

40µ

60µ

80µ

100µ

200 nm 25 nm

voltage [V]

curre

nt [A

]

local modification of sp2 structure by

current pulse

F. Kreupl et al. 12

Carbon Nanotubes In vacuum ~ 12 µA current possible

Jin et al., nature 18 nanotechnology | VOL 3 | JANUARY 2008 |

on-state off-state 12 uA

switch on 6 uA, 1.6V

on-state

F. Kreupl et al. 13

Allotropes of Carbon (investigated)

• Carbon Nanotubes - sp2-type - difficult to integrate - high conductivity

• Graphene or Conductive Carbon - sp2-type - easy to integrate

- high conductivity

• Insulating carbon - sp3-type, diamond-like

- easy to integrate - high resistivity

graphene layers

SiO2

nanotube

source: J. Robertson

F. Kreupl et al. 14

Conductive Carbon (CC)

• Conductive Carbon is – graphene-like – easily deposited (CVD) – can be used as interconnect

material (highly conductive) – easy to pattern

SiO2

conductive carbon in via

R. Seidel. et al., Chemical Vapor Deposition Growth of Single-Walled Carbon Nanotubes at 600 °C and a Simple Growth Model J. Phys. Chem. B, (2004) G. Aichmayr et al., Carbon-high-k Trench Capacitor for the 40nm DRAM Generation, VLSI Technology, (2007)

F. Kreupl et al. 15

Conductive Carbon: Memory Cell

• Carbon memory cells with varying diameter

30 nm

F. Kreupl et al. 16

Conductive Carbon: Critical Current

• Critical current density of 350 MA/cm2 observed • Appropriate cell diameter ~ 6 nm for I < 100 µA Use spacer, cladding or self-assembled nano-pores

F. Kreupl et al. 17

Conductive Carbon (CC): Switching

• Shmoo-plot of 40 nm diameter CC memory cell smaller diameter, current compliance and optimized

pulses required

1 2 3 4 5 6 7 8 9 10 11 12102103104105106107108109

1010

Reset 1 Set 1 Reset 2

Resi

stan

ce (O

hm)

Pulse Amplitude (V) 1 2 3 4103104105106107108109

Resi

stan

ce (O

hm)

Cycle

Reset Set

5 ns reset 300 ns set pulse

F. Kreupl et al. 18

Insulating Carbon (IC): Memory Cell

• Insulating diamond-like carbon film • First switching occurs now from high to low state

F. Kreupl et al. 19

Insulating Carbon (IC): Critical Field

• Quasi-static switching curves determine critical field • Switching power is about 50µW with leakage currents. • Very low power levels: 5 µA @ 1.5 V (P= 7.5 µW)

350 nm wide

150 nm wide

F. Kreupl et al. 20

Ins. Carbon: Read Endurance & Speed

• Read endurance at 75 degree C: 2.3*1013 read cycles at 0.1 V.

• Switching speed is faster than 11 ns.

0 100 200 300 400 500 6000

2

4

switching event

11 ns

puls

e he

ight

(V)

t (ns)10M 100M 1G 10G 100G 1T 10T0

25p50p75p

9µ

10µon-state

off-state

re-adjustmentof the probes

curre

nt [A

] @ 0

.1 V

number of read cycles

F. Kreupl et al. 21

Insulating Carbon: Switching

• I(V) curves show similar behavior after pulses • Resistance level can be trimmed by individual

voltage pulses (multi-level capability)

0.02 0.04 0.06 0.08 0.10

10p

100p

1n

10n

100n

1µ

after 5 V 500 µs after 5 V 100 ns after 5 V 500 µs after 5 V 100 ns

curre

nt [A

]

Voltage [V]0 1 2 3 4 5 6 7 8

6k8k

10k12k14k16k18k20k22k

w=100 ns w=5 ns

Resi

stan

ce [Ω

]

Voltage [V]

F. Kreupl et al. 22

Insulating Carbon: Filament size

crater with ~10 nm diameter

10 V pulse evaporates metal filament ~ 10 nm @ 10 V use carbon as current spreader

F. Kreupl et al. 23

Conclusions • New carbon memory proposed based on

sp2 to sp3 conversion • Inherently fast: reset ~ ns, set ~ ns • Nanotubes need ~30 µA @ 8V • Graphene-like Conductive Carbon

needs pores < 6 nm • Insulating Carbon shows lowest switching

power: 5 µA @ 1.5 V (P= 7.5 µW) • Pulses and cell design needs to be optimized • Should also work with Fullerens, Graphene and

Diamonds

F. Kreupl et al. 24

Acknowledgement Qimonda Fab Dresden Roland Thewes, Werner Pamler, Walter Weber, Uli Klostermann, Klaus-Dieter Ufert, Henning Riechert Stanford University Questions?

Thank you!