junction scaling technology for the sub 90nm node and beyond · 2004 wcjt meeting, portland, or ®...

®®2004 WCJT meeting, Portland, OR

Junction Scaling Technology for Junction Scaling Technology for the Sub 90nm Node and beyondthe Sub 90nm Node and beyond

Jack Hwang, Hal Kennel, Jack Hwang, Hal Kennel, Paul Paul PackanPackan , Mitch Taylor, , Mitch Taylor,

Mark Liu, Robert JamesMark Liu, Robert JamesIntel CorporationIntel Corporation


OutlineOutline

�� Introduction to transistor Introduction to transistor scalingscaling

�� Junction scaling constraintsJunction scaling constraints�� RTA developmentsRTA developments�� CoCo--doping effectsdoping effects�� ConclusionConclusion


MooresMoores LawLaw

M. Bohr, Intel Developer Form 2002M. Bohr, Intel Developer Form 2002


Scaling MOS TransistorsScaling MOS Transistors

�� MOS scaling theory MOS scaling theory requires scaling of all requires scaling of all geometric dimensionsgeometric dimensions

�� Active dopant Active dopant concentrations must be concentrations must be increased to maintain increased to maintain resistanceresistance

�� Limits are being pushed Limits are being pushed in this areain this area

DrainSource

Gate

DrainSource

Gate

Voltage Potential Contours


G. Marcyk, Tri gate announcement, Sept 17, 2002 ww w.intel.com/research/silicon

Junction depth Xj needs to scale with Lg 0.7X/node


Why Is High Activation Important?Why Is High Activation Important?

�� SDE resistance is a strong function of doping conce ntration SDE resistance is a strong function of doping conce ntration �� SDE resistance will increase significantly at junct ions below SDE resistance will increase significantly at junct ions below

35 nm.35 nm.

S. Thompson, VLSI Symp. 1998 T. Ghani, VLSI Symp. 2000

100

300

500

700

SDE Depth (nm)

SD

E R

esis

tanc

e

3E191E201E21

2.5nm/dec

10 20 30 40 50 60100

300

500

700

SDE Depth (nm)

SD

E R

esis

tanc

e

3E191E201E21

2.5nm/dec

10 20 30 40 50 60


Drive current and Drive current and RRextext

0.0

0.05

0.10

0.15

0.20

0.25

90

100

110

120

130

140

0 50 100 150 200 250Junction Depth (nm)

Lm

et@

1 nA

IO

FF

RE

XT

REXT

LMET

0.0

0.05

0.10

0.15

0.20

0.25

90

100

110

120

130

140

0 50 100 150 200 250Junction Depth (nm)

Lm

et@

1 nA

IO

FF

RE

XT

REXT

LMET

��As junction depths are decreased, short channel As junction depths are decreased, short channel effects improve showing gains in effects improve showing gains in IIdsatdsat for fixed for fixed IIoffoff..

��However, However, RREXTEXT increases very rapidly below 30increases very rapidly below 30 --40 nm 40 nm resulting in resulting in IIdsatdsat degradation.degradation.

Thompson et. al, Intel Technology Journal, Q3'98


OutlineOutline


�� Junction scaling constraintsJunction scaling constraints�� RTA advancesRTA advances�� CoCo--doping conceptsdoping concepts�� ConclusionConclusion


RTA and implant have been the traditional RTA and implant have been the traditional drivers for junction scalingdrivers for junction scaling

Known factors that Known factors that control control RRextext for 0.7X for 0.7X junction scaling:junction scaling:��Low energy implantLow energy implant��Faster RTPFaster RTP��CoCo--impurity additionimpurity addition��Junction abruptnessJunction abruptness��SDE SDE underdiffusionunderdiffusion

Junction targets for 65 nm is still TBDJunction targets for 65 nm is still TBD


Aggressive scaling at 65nm Aggressive scaling at 65nm allows for little allows for little dopantdopant diffusiondiffusion

�� 0.7X junction scaling 0.7X junction scaling at 65 nm node at 65 nm node requires 17 nm requires 17 nm XjXj..

�� XjXj for Implanted 0.5 for Implanted 0.5 keVkeV B11 implant is 13 B11 implant is 13 nm nm XjXj for 1E19 conc.for 1E19 conc.

�� How much farther can How much farther can XjXj be scaled?be scaled?

�� Other constraints?Other constraints?

Implanted B11+ 2E15 0.5 keV

1E+16

1E+17

1E+18

1E+19

1E+20

1E+21

1E+22

0 200 400 600 800 1000Depth (A)

Con

cent

ratio

n (a

t/cm

3)


No diffusion poses challenges No diffusion poses challenges for SDE for SDE underdiffusionunderdiffusion (XUD)(XUD)

�� SDE dimensions are scaled by 0.7X to meet SDE dimensions are scaled by 0.7X to meet LgateLgate target.target.�� Applying 0.7X scaling at 65 nm requires a 8nm XUD v s. Applying 0.7X scaling at 65 nm requires a 8nm XUD v s.

17 nm 17 nm XjXj --> Lack of diffusion may make it difficult to > Lack of diffusion may make it difficult to optimize optimize XjXj and XUD for transistor performance.and XUD for transistor performance.

T. Ghani, et al., VLSI symposium 2000


2D junction profile2D junction profile

Possible solutions for XUD• Improve junction abruptness• Decouple XUD from Xj. Xj is

not adequate to fully characterize the system.

2D dopant profile techniques can help assist the optimization of the junction.

S. Corcoran


OutlineOutline




RTA trend RTA trend -- faster is betterfaster is better

�� B B RsRs--XjXj benefits from spike anneal, but As is insensitivebenefits from spike anneal, but As is insensitive�� As (NMOS) may show limited benefits from future RTA As (NMOS) may show limited benefits from future RTA

developmentsdevelopments

B 1keV, 1e15As 4keV, 2e15

Kennel et al. IEDM 2002Kennel et al. IEDM 2002


Ramp rate up/down250/180400/1801000/180

200/80

1000C/s ramp and cool

Future RTA developmentsFuture RTA developments

Tool developments need to focus on improving Tool developments need to focus on improving cooldowncooldown ratesrates


0.7X Junction scaling 0.7X Junction scaling --> 0.7X uniformity > 0.7X uniformity scalingscaling

�� Simulated Simulated ∆∆∆∆∆∆∆∆T for % T for % XjXj control of B11 2.0E15 0.5 control of B11 2.0E15 0.5 keVkeV implant.implant.�� Any spatial temperature nonAny spatial temperature non --uniformity reduces number of die uniformity reduces number of die

with optimal performance.with optimal performance.�� The 65nm technology anneal solution will be evaluat ed based The 65nm technology anneal solution will be evaluat ed based

on performance and uniformityon performance and uniformity

0

10

20

30

40

50

60

50100150

Lg (nm)

Xj (

nm)

0

0.5

1

∆∆ ∆∆T

for

cons

tant

%

Xj (

arb

units

)

Junction Xj target

Spike anneal tolerance


OutlineOutline




Spike anneal and coSpike anneal and co --doping doping effectseffects

Spike anneals betterSpike anneals better�� Higher peak T @ Higher peak T @ XjXj leads leads

to improved solid solubilityto improved solid solubility�� Less TED (higher T)Less TED (higher T)

B+F coB+F co --implants show implants show significant significant RsRs--XjXjimprovement to spike improvement to spike anneal anneal

B 0.5k, 1e15, vary Tpeak

Co-doping effects can yield comparable Rs-Xjbenefits to RTA advances.

Kennel et al. IEDM 2002Kennel et al. IEDM 2002


Fluorine coFluorine co --implantimplant

�� Location of coLocation of co --dopantdopantrelative to junction is an relative to junction is an important parameter. important parameter.

�� Implant F deeper than B Implant F deeper than B for best results.for best results.

�� Fluorine coFluorine co --doping doping improves Junction improves Junction abrutpnesabrutpnes . .

�� Plateau is elevated as tail Plateau is elevated as tail is reduced.is reduced.

�� CoCo--doping effect is doping effect is chemical as chemical as SiSi PA control PA control experiment shows little experiment shows little impact on B11 profile.impact on B11 profile.

B:1e15, F:2e15

Review: Robertson IWJT 2001Review: Robertson IWJT 2001


Fluorine and Point DefectsFluorine and Point Defects

Si

Si

Si Si

Si

Si Si Si

Si

F

Kickout of Fluorineconsumes Ints

Trapping of F by Vac clustersduring regrowth

Mechanisms supported by Mechanisms supported by abab--initioinitio physical modeling physical modeling (Dunham SISPAD (Dunham SISPAD 2002, 2002, ShanoShano IEDM 2001)IEDM 2001)

Si

Si

Si Si

Si Si Si Si

Si

Si

F

F


Germanium coGermanium co --implants implants

�� Effect of high Effect of high GeGeconcentration is concentration is equivalent to PA for equivalent to PA for junction profile.junction profile.

�� SPE dominates coSPE dominates co --implant effect.implant effect.

�� Effect is physical, Effect is physical, not chemical.not chemical.

�� Addition of Addition of GeGe can can produce strain.produce strain.

1E+17

1E+18

1E+19

1E+20

1E+21

1E+22

0 500 1000 1500 2000 2500 3000

depth (A)

conc

entr

atio

n (a

t/cm

3)

No PAGe PA

Si PA

B11+ 1.0E16 5 B11+ 1.0E16 5 keVkeVSiSi PA 2.0E16 25 PA 2.0E16 25 keVkeVGeGe PA 2.0E16 50 PA 2.0E16 50 keVkeV


SPE effects dominate coSPE effects dominate co --implant effectimplant effect

Boron 1.0E16 5 keV

60

70

80

90

100

110

1065 1070 1075 1080 1085

Spike anneal temp(C)

Rs

(ohm

s/sq

)No PA

Si or Ge PA


Boron implant conditions can not Boron implant conditions can not completely completely amorphizeamorphize SiSi

Boron 1.0E16 5 keVBoron 1.0E16 5 keVGe-2.0E16 50 kev

Lack of complete amorphization has significant penalt ies for dopant activation (TEM is post anneal)


Defect engineeringDefect engineering

�� During low temperature During low temperature anneals, extended defects anneals, extended defects formform

�� These extended defects These extended defects may contain dopant atomsmay contain dopant atoms

�� The growth and The growth and dissolution of these dissolution of these defects determines the defects determines the dopant diffusion.dopant diffusion.

�� CoCo--doping effects can doping effects can modulate the evolution of modulate the evolution of these effects.these effects. dopant-interstitial

clusters

vacancies

interstitial clusters {311}

siliconinterstitials

dislocation loop


Are there any useful coAre there any useful co --doping doping interactions still to be discovered?interactions still to be discovered?

2 He

5 B

6 C

7 N

8 O

9 F

10 Ne

13 Al

14 Si

15 P

16 S

17 Cl

18 Ar

31 Ga

32 Ge

33 As

34 Se

35 Br

36 Kr

49 In

50 Sn

51 Sb

52 Te

53 I

54 X3

81 Tl

82 Pb

83 Bi

84 Po

85 At

86 Rn

More applications of other co-dopant applicationshave been documented in the literature.


Improving Improving dopantdopant solid solid solubilitysolubility

1E22

1E21

1E20800 1000 1200

Temperature ( oC)

Con

cent

ratio

n (/

cm3 )

Phosphorus

Arsenic

Active

Boron

Faster ramp rates and higher temperature to improve dopant solid solubility


Thermal History Thermal History

�� Subsequent low temperature Subsequent low temperature anneals nucleate defectsanneals nucleate defects

�� Extended defects can act as sinks Extended defects can act as sinks �� Process needs to be optimized for Process needs to be optimized for

full thermal process historyfull thermal process history

Si PA + RTP

Si PA + Furnace Anneal+ RTP

Si PA + RTP Preanneal + RTP

1E21

1E19

1E170 10 20 30 40 50

1E20

1E18

1E22

Depth (nm)

Con

cent

ratio

n (/

cm3)


ConclusionConclusion�� Aggressive scaling of Aggressive scaling of XjXj at 65 nm node will at 65 nm node will

present challenges in maintaining adequate XUD present challenges in maintaining adequate XUD for the SDE.for the SDE.

�� Future RTA hardware improvements will require Future RTA hardware improvements will require higher peak temperatures and faster higher peak temperatures and faster cooldowncooldownrates. rates.

�� Junction scaling efforts will require improvements Junction scaling efforts will require improvements in RTA temperature uniformity.in RTA temperature uniformity.

�� CoCo--dopantsdopants effects can be chemical or physical. effects can be chemical or physical. These effects can show comparable These effects can show comparable RsRs--XjXj benefits benefits to recent RTA advances.to recent RTA advances.

�� MooresMoores law will continue.law will continue.


AcknowledgementsAcknowledgements

Special thanks to the following people at Special thanks to the following people at Intel for their support:Intel for their support:

Craig Craig AndykeAndyke (Implant)(Implant)Andre Andre BudrevichBudrevich , Sean Corcoran (SIMS), Sean Corcoran (SIMS)Kevin Johnson (TEM)Kevin Johnson (TEM)Markus Kuhn (XRD)Markus Kuhn (XRD)

junction scaling technology for the sub 90nm node and beyond · 2004 wcjt meeting, portland, or ®...

Documents