Accelerating the next technology revolution

Copyright ©2010 SEMATECH, Inc. SEMATECH, and the SEMATECH logo are registered servicemarks of SEMATECH, Inc. International SEMATECH Manufacturing Initiative, ISMI, Advanced Materials Research Center and AMRC are servicemarks of SEMATECH, Inc. All other servicemarks and trademarks are the property of their respective owners.

Memory Overview and RRAM Materials Development at SEMATECH

Paul Kirsch,Director Front End ProcessesAugust 22, 2010

Nonvolatile Memory SeminarHot Chips ConferenceMemorial AuditoriumStanford University

5 August 2010 2

Outline

• NVM Memory Trends & Challenges

• Technology Space for RRAM

• Challenges and SEMATECH RRAM Results

High density, low cost, NVM (NAND) Y. N

ishi

, S.M

. Sze

, S

tanf

ord

Uni

v (s

ourc

e IT

RS

200

0)

April 2, 2010 3

System on Chip (SOC)Mobile

Dense, local memory – improve system performance

Future trends: Memory for mobile and SOC applications

• IMPACT: High density, low cost NVM important for mobile application

• IMPACT: Local (embedded) memory important for System on Chip

5 August 2010 4

Advantage Challenges“Speculative”

PriorityPhase Change RAM • Speed & cost vs. NOR • Density, Power (Ireset) production

SiN / Nano Crystal Charge Trap

• Relatively mature• difficult to beat floating gate• candidate for 3 -D

1

RRAM • High speed & density • Materials, I reset, Reliability 2

STT-RAM • endurance, speed , power• Magnetic domain size effect • Etch difficulty• Challenging materials

2

Mechanical memory• Power• Leakage

• Reliable operation• Scalability

4

Molecular memory • Density (Molecule size ~nm) • Poor thermal stability 5

Vertical String • 2-5x lower $$ vs. stacked • deposited tunnel ox• transistor in trench sidewall

1

Cross bar array • effective density below 4F 2• Litho $: more costly vs. NAND• selector device

2

Stacked • similar to current NAND • Litho $: more costly vs. NAND 3

Cel

lA

rchi

tect

ure

Sub 20nm memory cell, architecture candidates

5 August 2010 5

[1] W. Y. Choi, and T.-J. King Liu IEDM p. 603 (2007).[2] A. Driskill-Smith, Y. Huai Future Fab p. 28 (2007).[3] J. E. Green Nature v. 445 p. 414 (2007).[4] B. Yu IEEE Trans on Nanotech 7, p. 496 (2008).[5] Kryder, et. al. IEEE TRANS ON MAGNETICS, 45, NO. 10, (2009)

• MLC = NAND• Cost structure = NAND• Function, reliability = NAND• Density > NAND • Speed > NAND• RRAM, STT interesting.

0 10 20 30 40 50100

101

102

103

104

105

NW-PC

STT

NanowireMolecular

DRAM

PCFeRAM

NEMS SRAM

MRAM

NOR

low power (fJ/bit) mid power (pJ/bit) mid-hi power (low nJ/bit) high power (nJ/bit)

Switc

hing

Tim

e(re

ad+w

rite)

[ns]

Density AF2

NAND

RR

Sub 20nm memory benchmarking

Success Criteria: Future NVM

BL1

BL2

BL3

WL1 WL2 WL3

1F Unit cell = 4F 2

small A = small cell

Bit

cost

($/G

b)

Which space should RRAM target?Associated challenges to compete with DRAM/NAND/NOR

Program/write speed (ns)

DRAM

NAND

NOR

SRAM

0.01

0.1

1

10

100

101 102 103 104 105 106 107

HDD

3-D multilayer crossbar(dense NVM)

1T1R BEOL memory (SOC memory)

To achieve 3-D CrossBar:• $: 9 critical levels

in 4 layer stack• EUV?• DP?• Selector• Etch

25nm L/S

To Compete w/NAND• Cost • Reset current• MLC (density)• 3-D integration

To compete w/DRAM:• Better endurance• Better reset current• Current scaling• Uniformity

RRAM challenge: compete w/ NAND cost

April 2, 2010 6

RR

AM

: 10ns reported

Challenges with RRAM

• Too many materials (manufacturability?)

• Reducing reset currents (power)

• Uniformity

• Endurance

• Integration with selector device (1T1R, 1D1R)

5 August 2010 7

5 August 2010 8

Many RRAM materials: Many not manufacturable

H

Li

Na

K

Rb

Cs

Fr

Be

Mg

Ca

Sr

Ba

Ra

Sc

Y

La

Ac

Ti

Zr

Hf

Rf

V

Nb

Ta

Db

Cr

Mo

W

Sg

Mn

Tc

Re

Bh

Fe

Ru

Os

Hs

Co

Rh

Ir

Mt

Ni

Pd

Pt

Ds

Cu

Ag

Au

Rg

Zn

Cd

Hg

B

Al

Ga

In

Tl

C

Si

Ge

Sn

Pb

N

P

As

Sb

Bi

O

S

Se

Te

Po

F

Cl

Br

I

At

Ne

Ar

Kr

Xe

Rn

He

Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu

Th Pa U Np Pu Am Cm Bk Cf Es Fm Md No Lr

• Many materials risky to put into a semiconductor development line

X = Materials in RRAM literature report

5 August 2010 9

Which RRAM materials are manufacturing worthy?

H

Li

Na

K

Rb

Cs

Fr

Be

Mg

Ca

Sr

Ba

Ra

Sc

Y

La

Ac

Ti

Zr

Hf

Rf

V

Nb

Ta

Db

Cr

Mo

W

Sg

Mn

Tc

Re

Bh

Fe

Ru

Os

Hs

Co

Rh

Ir

Mt

Ni

Pd

Pt

Ds

Cu

Ag

Au

Rg

Zn

Cd

Hg

B

Al

Ga

In

Tl

C

Si

Ge

Sn

Pb

N

P

As

Sb

Bi

O

S

Se

Te

Po

F

Cl

Br

I

At

Ne

Ar

Kr

Xe

Rn

He

Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu

Th Pa U Np Pu Am Cm Bk Cf Es Fm Md No Lr

• Focus on engineering materials that are already in the semiconductor fab (green)

X = Materials in RRAM literature report

X = Materials in fabs today

5 August 2010 10

STT-MRAM Materials also challenging

H

Li

Na

K

Rb

Cs

Fr

Be

Mg

Ca

Sr

Ba

Ra

Sc

Y

La

Ac

Ti

Zr

Hf

Rf

V

Nb

Ta

Db

Cr

Mo

W

Sg

Mn

Tc

Re

Bh

Fe

Ru

Os

Hs

Co

Rh

Ir

Mt

Ni

Pd

Pt

Ds

Cu

Ag

Au

Rg

Zn

Cd

Hg

B

Al

Ga

In

Tl

C

Si

Ge

Sn

Pb

N

P

As

Sb

Bi

O

S

Se

Te

Po

F

Cl

Br

I

At

Ne

Ar

Kr

Xe

Rn

He

Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu

Th Pa U Np Pu Am Cm Bk Cf Es Fm Md No Lr

X = Materials in STT-MRAM

11

PVD metal oxide RRAM development

• PVD ideal for Stoichiometry Control, thin, uniform films

• ALD films also developed, but may still need PVD for electrode.

• IMPACT: Correct material Me:O enables correct function.

O c

onte

nt in

RR

AM

(x

in M

eOx)

(a.

u.)

StoichiometricControl

Stoichiometriesof interest

Process Condition

5 August 2010

• Resistance switching currents (I reset) improved ~ 1000x.

• I reset improved to 1~10 µA range – desire 10x more improvement.

• IMPACT: Sub 1 µA device likely meets U.S. DARPA target for Switching Energy (fJ/bit)

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.510-9

10-8

10-7

10-6

10-5

10-4

10-3

10-2

10-1

Set

/ R

eset

Cur

rent

[A]

VD [V]

1 2 3 4 5 6 7 8 9 10

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5

VD [V]

1 2 3 4 5 6 7 8 9 10

-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5

VD [V]

1 2 3 4 5 6 7 8 9 10

forming

1

2

4

3

0

Progress in reducing reset currents

12

Reducing Metal oxide RRAM Reset Currents

5 August 2010 13

• SEMATECH has exceeded member company target for 2010 Ireset

• 5 micro-amp reset currents achieved with manufacturable metal oxide

0

5

10

15

20

# M

ater

ials

S

cree

ned

ALD PVD

Process Type

Target

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

10 100 1000 10000 100000

Z. Wei, IEDM 2008 p. 239

LRS HRS

Resistance (Ohm)

Pro

babi

lity

RRAM Reset Uniformity Improves

• Uniformity of High Resistance state improve across wafer (>300pts).• IMPACT: Uniformity is key issue for Manufacturability - Multilevel Cell

Improvement

5 August 2010 14

5 August 2010

-2 -1 0 1 210-9

10-8

10-7

10-6

10-5

10-4

10-3

10-2

100001 cycle100002 cycle100003 cycle100004 cycle100005 cycle100006 cycle100007 cycle100008 cycle100009 cycle

Voltage [V]C

urre

nt [A

]

After 105 cycles

-2 -1 0 1 210-9

10-8

10-7

10-6

10-5

10-4

10-3

10-2

100001 cycle100002 cycle100003 cycle100004 cycle100005 cycle100006 cycle100007 cycle100008 cycle100009 cycle

Voltage [V]C

urre

nt [A

]

-2 -1 0 1 210-9

10-8

10-7

10-6

10-5

10-4

10-3

10-2

100001 cycle100002 cycle100003 cycle100004 cycle100005 cycle100006 cycle100007 cycle100008 cycle

-2 -1 0 1 210-9

10-8

10-7

10-6

10-5

10-4

10-3

10-2

100001 cycle100002 cycle100003 cycle100004 cycle100005 cycle100006 cycle100007 cycle100008 cycle100009 cycle

Voltage [V]C

urre

nt [A

]

After 105 cyclesDC checkBetween 1st and 2nd 105 cycles

• Good initial RRAM cycling (2x10 5 +) – much work to do yet• IMPACT: If RRAM achieves fJ/bit and 10 15 cycles it competes with DRAM

100 101 102 103 104 10510-7

10-6

10-5

10-4

10-3

10-2

Vread=0.5V

Number of Cycles

Cur

rent

[A]

100 101 102 103 104 105

Number of Cycles

Vread=0.5VAfter 1e5 cycles

100 101 102 103 104 10510-7

10-6

10-5

10-4

10-3

10-2

Vread=0.5V

Number of Cycles

Cur

rent

[A]

100 101 102 103 104 105

Number of Cycles

Vread=0.5VAfter 1e5 cycles

Initial 105 cycles 2nd 105 cycles

Encouraging RRAM endurance check

15

• Further work needed on devices meeting Ireset, but encouraging initial data.

Metal oxide RRAM has demonstrated reasonable retention

100 101 102 103 104

1E-4

1E-5

1E-6Cur

rent

(A

)

April 2, 2010 16

Selector Challenge

Reduce RRAM Reset

Currents <1µA

Relax Selector

Drive Needs <106 A/cm2

5 August 2010 17

Selector Device

Bitline

Wordline

RRAMElement

• Why does reset current matter?

• SELECTOR lesser Ireset means lesser diode current requirements

• IMPACT: Selector flexibility eases challenge of array integration

D. Kau, Intel, IEDM 2009

1D1R

Trade-off

5 August 2010 18

O

OO

O

O

O

metal

metal

O

Basic metal oxide RRAM Mechanism

• Switching, retention, endurance controlled by defec ts

• IMPACT: Need to understand details for reliable product

Binary oxides : filament formation

OO atom

O vacancy

Conductive Filament

“Good” Insulator

SEMATECH test structureTest structure allows reproducible electrical / reliability data

Test Structure

Structure Purpose

1. Contact -type MIM structure

(x-section)

• Quick materials studies

2. Stack -type MIM

structure(X-section)

• Quick materials studies

3. 1T1R test structure(top view)

• Iso 1T1R, Iso 1R, • Cross Pt Array1T1R

• Detailed Ireset studies• Detailed parasitic studies• Array retention, endurance• 25nm x 25nm CD

Si

SiO2

Bottom electrode

TiNMetal cap

Top electrode

Metal Oxide

Substrate

Insulator

TEMOx

Bottom electrode

ILD

April 2, 2010 19

Overview of SEMATECH capabilities

State of Art Test Structure Design

Process Design

New Memory Materials

Sub 30nm Test Structure

Fabrication

Reliability & Parametric

Test

Transfer material, flow, equipment to

manufacturers

5 August 2010 20

Partner with universities

Partner with CNSE,SVTC

Partner with Universities

Partner with Mfg Companies

SEMATECH 1T1R Test Pattern Design

SEMATECH 20nm Nano Devices

SEMATECH Mfg Members Use Modules

5 August 2010 21

Conclusions

• Planar floating gate will be pushed to 2X nm.

• RRAM may compete as SOC memory (BEOL).

• RRAM may fit between NAND &DRAM cost space (stand alone)– Cost will be key– 4F2 cell with ~4 levels achieve effective 1F2 (MLC necessary)

• Materials challenges addressed– Manufacturable MeOx and electrode– Reset / set currents are approaching acceptable levels (1micro-amp)– Improved Uniformity Encouraging– Endurance, retention are promising, need improvement >1e5 cycles– Selector device and polar / bipolar design are challenges

• Quality test structures (1T1R) assist in accurate characterization, development