Modeling of Hot-carrier Degradation: Physics and ... · Employment of LDD structures V ds is scaled...

Institute for Microelectronics

Vienna University of Technology

http://www.iue.tuwien.ac.at

Modeling of Hot-carrier Degradation: Physics and Controversial Issues

S.E. Tyaginov

2

Acknowledgement

This work would not have been possible without the support of:

My Colleagues at TU Wien:

I. Starkov, O. Triebl, D. Osintsev, M. Bina, and T. Grasser

Industrial Partners From AMS:

H. Enichlmair, J.M. Park, and R. Minixhofer

RWTH Aachen

Prof. Ch. Jungemann

Colleagues from IMEC:

J. Franco and B. Kazcer

Colleagues from the Ioffe Institute, Russia

M.I. Vexler and I.V. Grekhov

ISEN-IM2NP

Prof. A. Bravaix

3

Outline

Hot-carrier degradation: what is it?

Main peculiarities

Physical picture behind hot-carrier degradation

Physics-based modeling of HCD

The modeling paradigm which integrates:

Carrier transport

Microscopic mechanisms of defect creation

Modeling of degraded devices

Controversial/open issues

Conclusions

4

HCD Basics: Definition

Hot-carrier degradation (HCD)

Build-up of defects

At/near the Si/SiO2 interface

Field acceleration → gained sufficiently high energy → “hot” carriers

Interface states → density Nit

Nit is a distributed quantity

Varies with the coordinate along the interface

And in energy

5

HCD Basics: Nit

Interface states

Can capture electrons/holes

Become charged

Perturbs the electrostatics

Result in a threshold voltage shift ΔVth

6

HCD Basics: Nit

Interface states

Act as additional scattering centers

Thereby degrading:

Mobility

Transconductance ΔGm

Linear drain current ΔIdlin

etc

7

HCD Basics: What about Not?

Oxide traps with the density Not: contribute?

Trapping/detrapping in the oxide bulk:

Plays a crucial role in BTI

BTI and HCD: similar microscopic origin

Therefore, bulk oxide traps are expected in HCD

Responsible for

SILC

And for TDDB

T. Grasser et al., TED, 2011

D. Varghese et al., EDL, 2005

H. Park et al., IIRW, 2008

R. O’Connor et al., IRPS, 2008

8

HCD Basics: What about Not?

Turn-around effects in HCD:

These effects are due to the partial compensation of the:

Charge stored in oxide traps

By interface state trapped charge

9

HCD Basics: What is the Driving Force?

Field-driven paradigm

Lucky Electron Model

The main assumptions is that the “lucky-electron”:

Has energy enough to overcome the barrier

Impinges the interface without collisions

And without being scattered back

This energy is gained by the electric field:

φit – the potential barrier

λ – the electron free path

Em is the maximum electric field

C. Hu et al., TED, 1985

n

Eqdit

miteW

ItCN

/

10


Measures to suppress HCD In 80s device linear dimensions were reduced rather quickly

Accompanied by a slower supply voltage scaling

Result:

High electric fields

And severe hot-carrier degradation

Strategies:

Supply voltage has to reduce faster vs. device dimensions

Constant field scaling

Introduction of LDD structures

T-.Y. Huang et al., IEDM, 1985

T. Hori et al., TED, 1992

11


Energy deposited by carriers is the driving force

The IBM group investigated:

Fowler-Nordheim, direct tunneling stresses

Substrate hot-electron/hole stresses

Channel hot-electron/hole stresses

Interface state generation probability

Depends only on the carrier energy

Not on the field

Insensitive to stress mechanisms

Energy deposited by carriers

drives degradation

D.J. DiMaria, S.W. Stasiak, JAP, 1989

D.J. DiMaria, J.H. Stathis. JAP, 2001

12


Possible driving forces

Electric field, carrier average energy

Are distributed quantities

With a maximum close to the drain end of the gate

Hints that HCD is a strongly localized phenomenon

13

Outline


Main peculiarities




Carrier transport




Conclusions

14

Main Peculiarities: Localization

Strong localization of HCD

Interface state density lateral profile Nit(x)

Extracted from charge-pumping (CP) data

For two different oxide thicknesses

Nit(x) profiles feature peaks

Near the electric field maximum

M.G. Ancona et al., TED 1988

15

Main Peculiarities: Carrier Ensemble

Carrier ensemble, not a solitary carrier

Si-H bond-breakage is a stochastic process

Carrier with a certain energy → certain probability to break a bond

Carrier packet has substantially different energies

Each carrier/energy → a certain contribution to HCD

Evolution of the ensemble along the interface

Macroscopic quantity which

Describes the cumulative ability

to rupture Si-H bonds

Different quantities used as driving force:

Different lateral distributions

Different Nit profiles

16

Main Peculiarities: From Hot to Cold Carriers

HCD suppression by a proper supply voltage scaling?

Device dimensions were reduced rather quickly

Slower scaling of power supply

High electric field in the MOSFET channel

Carrier acceleration → high energy → Si-H bond rupture by solitary carrier

Measures to suppress carrier heating:

Fast scaling of power supply

Employment of LDD structures

Vds is scaled down to 1.0-1.5 V

Threshold energy for Si-H dissociation: 3.0-3.5 eV

Halt of degradation? NO!

K. Hess et al., Circ. Dev. Mag., 2001

A. Bravaix et al., IRPS-2009

17


Energy exchange mechanisms Populating high energy fraction of the ensemble

Impact ionization

Auger recombination

Electron-phonon scattering

Electron-electron scattering

Is of special importance in

Ultra-scaled devices

K. Hess et al., Circ. Dev. Mag., 2001 A. Bravaix et al., IRPS-2009

18


Si-H bond-breakage mechanisms

Ultra-scaled devices Electron-electron scattering

Change of the dominant dissociation mechanism:

Single-particle (SP) → multiple-particle (MP) process

Long-channel/high-voltage devices

Carriers are rather hot

Bond rupture in a single collision

SP-process

Scaled devices

Several “colder” particles → subsequently bombard a bond

Bond excitation → rupture

MP-process

Interplay between SP- and MP-mechanisms

Change of the HCD worst-case conditions

19

Main Peculiarities: Worst-Case Conditions

Long-channel/high-voltage n-MOSFET Corresponds to the maximal substrate current Isub

Maximal impact ionization rate

Vgs = (0.4-0.5)Vds

Long-channel/high-voltage p-MOSFET Worst-case conditions (WCC) ↔ the maximum gate current Ig

Empirical link between the voltages is not established

Scaled devices: A single carrier is unlikely to trigger an SP-process

Carriers contributing to the MP-mechanism:

Require only a low energy

A large number of carriers

Carrier flux rather than the single-carrier energy becomes important

WCC ↔ Vds = Vgs

For both scaled n- and p-MOSFETs

20

Main Peculiarities: Temperature Behevior

HCD becomes less pronounced at elevated temperatures Contrary to BTI

An n-MOSFET was stressed at WCC

Threshold voltage shift due to HC stress

∆Vth(t) is less pronounced at higher T

This traditional tendency is typical only for long-channel devices

Ultra-scaled MOSFETs HCD becomes more significant at higher T

Dominant role of electron-electron scattering

Impacts the carrier ensemble

F.-C. Hsu and K.-Y. Chu, EDL,1984.

21

Main Peculiarities: Conclusions

Important: how carriers are distributed over energy

In a particular point in the device

By the carrier energy distribution function (DF)

Interplay between hotter and colder carriers


Strong localization of HCD ↔ driving force ↔ carrier transport

Temperature behavior ↔ scattering mechanisms ↔ carrier transport

Carrier transport

Intimately related to hot-carrier degradation

It is an essential fragment of the whole physical picture

↔ carrier transport

22

Outline


Main peculiarities




Carrier transport




Conclusions

23

Physical Picture: Carrier Transport

The carrier ensemble moved from the source to drain

Carriers are being accelerated by the electric field

Experience scattering events

Evolution of the carrier ensemble

The ensemble is described

By the energy distribution function

Probability to find a particle within [E; E+dE]

24


The evolution of the DF with the coordinate x

Source: carriers in equilibrium → Maxwellian distribution

Drain end of the gate: the carrier distribution is severely non-Maxwellian

Near the Nit peak: high-energy tails, plateau

Drain: again Maxwellian

5V n-MOSFET

standard 0.35 μm process

channel length: 0.5μm

25


Main scattering mechanisms in MOSFETs

Electron-phonon scattering

Scattering at ionized impurities

Impact ionization

Surface scattering

Auger recombination


In degraded devices: scattering on charged interface states

26


Carrier ensemble

Gains energy from the electric field

Loses and intermixes energy

Due to scattering

High-energetical fraction is depopulated

High-energy tail of the DF is distorted

Temperature behavior of HCD

Elevated temperatures

Scattering mechanisms are reinforced

High-energy tails of the DF is suppressed

“Hot” carriers are “frozen out”

27

Physical Picture: Vague Issues [not really]

We focused on the high-energetical fraction

How to distinguish between “hot” and “colder” carriers?

Why are we primarily interested in hot carriers?

Why is HCD observed even in ultra-scaled devices?

28

Physical Picture: Hydrogen in MOSFETs

Hydrogen is intentionally incorporated into CMOS devices

Si/SiO2 interface is non-regular

Dangling bonds

Pb centers

The electrostatics are disturbed

The mobility is reduced

Remedy: post-grow anneal

Hydrogen passivates dangling bonds

C.R. Helms, H.E. Poindexter, Rep. Prog. Phys, 1994

29

Physical Picture: Si-H Bonds

Si-H bond dissociation

Leads to electrically active centers

Contribute into HCD and (N/P)BTI

Si-H bond energetics are important

Ab initio calculations using density functional theory

Si-H dissociation reaction pathway:

Bonded hydrogen → transport state

Potential barrier: ~2eV

Portion of energy delivered to H: 2eV

High-energetical fraction of the ensemble

With energies above 2eV

To trigger the Si-H bond dissociation

Hot carriers!

K. Hess et al, Physica E 3, 1998

B. Tuttle and Ch.G. Van de Walle, PRB, 1999

30

Physical Picture: Si-H Bonds

Disparity: electron mass and the mass of the H nucleus

Difference ~1000 times

Total moment conservation

Direct bombardment: negligible portion of energy transferred

Bond dissociation is unlikely

Most probable pathway

Excitation of one of the bonding electrons to an antibonding state

A repulsive force acting on the H atom is induced

Followed by H release

K. Hess et al., Physica E 3, 1998

W. McMahon et al., Int. Conf. Mod. Sim. Micro, 2002

W. McMahon et al., IEEE Trans. Nanotech., 2003

31

Physical Picture: HCD in Scaled Devices

Is there any HCD?

The supply voltage in scaled devices is 1.0-1.5 V

Carriers with energies above the threshold for Si-H dissociation?

Unlikely!

Isn’t HCD still a concern?

Pioneering paper by Mizuno et al

Lch = 0.12μm

Stressed at WCC

Vds = 1.5V, Vgs = 0.7V

Severe linear drain current degradation

T. Mizuno et al., IEDM, 1992

32

Physical Picture: HCD in Scaled Devices

Factors responsible for HCD in scaled devices:

Energy exchange mechanisms

Populating high energy fraction of the ensemble

Single-particle process of Si-H bond dissociation

33

Physical Picture: Scattering Mechanisms

Impact ionization induced high-energy tails

Gate currents Ig is measure of high-energy tails

n-MOSFET (0.1μm process) → investigated using a Monte-Carlo simulator

Mechanism responsible for Ig:

Impact ionization feedback through

The vertical fields of the drain-substrate junction

High-energy tails of the DF: pronounced

J.D. Bude, VLSI Techn. Dig, 1995

34


The role of Auger recombination

n-MOSFETs: Lch = 0.3μm, dox = 40nm → tunneling is suppressed

Stressed: at Vds = 1.4-1.6V and T = 77, 300 K

Gate current Ig and the threshold voltage shift ΔVth

Ig is a criterion of hot-carriers

And carriers are hot

Physical mechanism

Vds ≥1.4 V → II contributes

Drain concentrations:

n = 1020 cm-3, p = 5·1014 cm-3

Recombination cannot be neglected

Auger recombination:

Two recombining carriers give their energy electron

Because n >> p

B. Ricco et al., IEDM, 1984

35


Electron-phonon interactions

Electrons can gain energy from phonons if

Number of absorbed phonons exceeds

Number of emitted phonons

DF expands beyond |e|Vds

This scenario was supported

By Monte-Carlo simulations

For an n-MOSFET, Leff = 0.1 μm

A. Abramo et al., IEDM, 1995

36


Electron-electron scattering (EES)

Is of special interest in nano-scale devices

1D Boltzmann transport equation with EES is included

Energies available from the electric field: |e|Vds

Calculated DFs for Vds = 0.5, 1.0, 1.5V

With and w/o EES

EES dramatically changes

the shape of the DF

DF propagates deeper than |e|Vds

P.A. Childs, C.C.C. Leung, JAP, 1996

M. Fischetti, S.E. Laux, IEDM, 1995, A. Ghetti et al, IEDM, 1998

37

Physical Picture: Multivibrational Excitation

Multivibrational excitation (MVE)

First success of the concept

H/D desorption

Bombardment by electrons

Tunneling from STM tip

D-passivated surfaces: more resistant

Difference in depassivation rates: 2 orders

Giant isotope effect

B.N.J. Persson, Ph. Avouris, Surf. Sci.,1997

J.W. Lyding et al., Appl. Surf. Sci., 1998

K. Stokbro et al., PRL, 1998

38


Si-H bond

Treated as a truncated harmonic oscillator

Eigenstates in the quantum well

Dissociation:

H hopping: last bonded → transport state

Vice versa = passivation

Electron flux

Phonon absorbtion → bond heating

With the rate Pu

Desorption → MVE decay

With the rate Pd

39


Explanation of the giant isotope effect

Biswas and collaborators simulated vibrational excitations

Using tight-binding molecular dynamics

Two types of vibrational modes:

Stretch mode

Very stable for both Si-H and Si-D

Up to 0.8 ns

Bending mode

Si-H: no decay within 0.8 ps

Si-D: bending mode

Dumps down within 1-2 oscillations

Si/SiO2 interfaces:

H emission via bend-bending distortion

Giant isotope effect is explained

R. Biswas et al., PRB 1998 R. Biswas et al., APL, 1998

40

Outline


Main peculiarities




Carrier transport




Conclusions

41

Physics-Based Models: Hess Model

The main breakthrough of the Hess model

Introduction of two competing mechanisms

SP- and MP-processes

Dominating HCD in

Long-channel/HV devices (SP-process)

Scaled devices (MP-process)

Related to

“Hot” (SP-process)

And “colder” carriers (MP-process)

HCD is controlled by the carrier distribution function

Or by another quantity derived from the DF

The carrier acceleration integral (AI)

K. Hess et al., Physica E, 1998


W. McMahon et al., IEEE Trans. Nanotech., 2003

42


The acceleration integral Desorption rate via: the bonding electron → antibonding state

F(E) is the carrier impact frequency

Per unit area

Per unit energy

σ(E) is the scattering cross section

P(E) is the probability of the desorption

Eth is the threshold energy for scattering

Flux F(E):

F(E) = f(E)g(E)v(E)

f(E) is the carrier DF, g(E) the DOS, v(E) the carrier velocity

K. Hess et al., Physica E, 1998 W. McMahon et al., Int. Conf. Mod. Sim. Micro, 2002

dEEPEEFIEth

e )()()(

43

Hess Model: Multivibrational Mode Concept

The MP-process The Si-H bond is treated as

Truncated harmonic oscillator

1/we – phonon life-time

– distance between the levels

Occupation number: Bose-Einstein

Phonon absorbtion = bond heating (Pu)

Desorption = multivibrational mode decay (Pd)

/

)/exp(exp1

BE

LBed

eu

LB

dB

MPTkwP

wP

TkP

ER

K. Hess et al., Physica E, 1998

dEEfEEFP

dEEfEEFP

phemi

Eth

u

phab

Eth

d

1)()(~

1)()(~

The MP-process rate:

44


Contribution of all levels

Not only SP- and MP-mechanisms

SP-process: excitation from the ground state

MP-process: from the last bonded state

Rate equations are simplified

Linked to the drain current Id

B.N.J. Persson, Ph. Avouris, Surf. Sci., 1997


edd

LBedu

wdEIP

TkwdEIP

)/exp(

lN

i

dd

i

i

evd

evd

fIAwfI

kTwfI

R0

exp

defines the population

of the i-th level

H hopping

45


Activation energy Ea for the Si-H bond-breakage

Statistically distributed

This is supported by ab initio calculations (DFT)

Dispersion of Ea → different power-law slopes:

21 )/(1)/(1~

2

2

1

1

t

p

t

pNit

2

2/1,

2/1,2/1,

2/1,

2/1,2/1,

2/1,

exp1

exp1

a

aam

a

aam

a

it

EE

EE

N

Two distributions with different:

Mean values Ea,1/2

Standard deviations σa,1/2

experimentally observed

K. Hess et al., Physica E 3, 1998;B. B.

Tuttle, PRB, 1999

46


H or D annealing of the dangling bonds?

Multivibrational mode concept

Giant isotope effect

MOSFETs with the Si/SiO2 interface passivated

By hydrogen

By deuterium

Subjected to hot-carrier stress

Devices with D-annealed interface

Demonstrate improved robustness

E.g. the transconductance degrades less

J.W. Lyding et al., APL, 1996

47


Advantages

A lot of pioneering concepts

Shortcomings Interface traps: microscopic level

Unconnected to the device level

Device life-time:

Time when Nit = Nit,critical

Degradation of such parameters as Gm, Idlin, etc:

Not really addressed

Necessity of the DF evaluation is acknowledged

But not incorporated into the approach

48

Physics-Based Models: Penzin Model

Adaptation of the Hess model for TCAD simulators Phenomenological simplification

Microscopic level is sacrificed

The bond rupture → described by kinetic equation:

Ea depends on: Hydrogen density

Transversal F┴ component of the electric field

The vicinity of the interface → capacitor

Released H and dangling bonds = charges → electric field

Prevents hydrogen ions from leaving the system

HC

HCHCH

HLBa

Ik

kTkEkk

nNkndt

dn

1

)/exp(

)(

0

0

n –concentration of passivated bonds

N0 – total bond concentration

k, γ – forward, backward reaction rate

k0 – attempt rate

kH – hot-carrier acceleration factor

IHC – local hot-carrier current (?)

δHC, ρHC – fitting parameters

F

nN

nNTkFEE LBaa

1

ln)0(

0

00

O. Penzin et al., TED, 2003

49

Physics-Based Models: Penzin Model

Activation energy dispersion

Captured by the model

Represents sublinear slope

Shortcomings

Carrier transport (?)

The hot-carrier acceleration factor (?)

“Local hot-carrier current”(?)

Criterion to distinguish “hot” and “cold” carriers

Based on the DF

Information about the Nit profile is hardly achievable

Kinetics of the trap generation and the device characteristics:

Are linked?

Rigorous comparison experimental/simulated device characteristics?

O. Penzin et al., TED, 2003

50

Physics-Based Models: Reaction-Diffusion (RD)

NBTI and HCD are related to Si-H bond-breakage

Differing only in the driving force

HCD + NBTI → united within RD

HCD, NBTI: diffusion-limited

Some of experimental observations:

NBTI: ~ t1/4

HCD: ~ t1/2

NBTI is a 1D problem

HCD is a 2D phenomenon

non-uniform Nit(x)

2/1)(

4/1)(

)(

0

)0(2/1)0()(

2/1

)(

0

)0(2/1)0()(

)(~

)(~

)()12/()/(1)2/(

)()2/1()/(1)/1(

2/1

2/1

tDN

tDN

tDNArdrtDrNAN

tDNdrAtDrNAN

H

HCD

it

H

NBTI

it

H

tD

HdHHd

HCD

it

H

tD

HdHHd

NBTI

it

H

H

H. Kufluoglu and M. Alam, Journ. Comput. Electron., 2004.

51

Physics-Based Models: Reaction-Diffusion

NBTI, HCD: diffusion limited

Stress is removed → recovery occurs rather quickly

NBTI: reaction limited

HCD: the recovery is rather weak

if there is any recovery at all

Model does not rely on carrier transport

Driving force behind the trap generation?

Nit distribution?

Localized nature of the damage

Not addressed

52

Physics-Based Models: Energy-Driven Paradigm


Of special interest in scaled MOSFETs

Supply voltage: very low

SP- mechanism is suppressed

EES populates the high-energy tail

Hump in the carrier DF

SP-contribution is increased

EES defines the temperature behavior

Acceleration of HCD at elevated temperatures

In extremely-scaled MOSFETs

P.A. Childs, C.C.C. Leung, JAP, 1996

S.E. Rauch and G. LaRosa, TDMR, 2001

S.E. Rauch et al., EDL, 1998

53


Driving force of HCD is the energy deposited by carriers

Not the maximal electric field

Beyond the 180 nm node

II rate/ Nit creation rate:

f(E) is a strongly decaying function

S(E) grows as a power-law

Trade-off results in:

Maximum at a certain energy

“Knee” energy

Weak function of Vds

dEESEf )()(f(E) – carrier DF

S(E) – reaction cross section

S.E. Rauch et al., EDL,1998

S.E. Rauch et al., TDMR, 2001

S.E. Rauch, G. LaRosa, IRPS, 2005

54


Main advantages

Substantially simplifies modeling of HCD

Time-consuming calculations of the DF → Avoided!

Empirical parameter

Disadvantages

Maximum is not necessarily narrow

Width: 1.5 - 2 eV

Dominant energy?

The concept does not deal with Nit as a distributed quantity

Strong localization of HCD → not captured

Similar to the Hess approach:

Life-time: interface state generation rate

ΔGm, ΔIdlin → critical value

55

Physics-Based Models: Bravaix Model

Main features from the Hess and the Rauch/LaRosa models


Idea: damage is defined by the carrier DF

Calculations of the DF → condition-related empirical factors

Rates in the Hess model

In the Bravaix model:

SMP is a fitting factor

Representing the reaction cross section

edd

LBedu

wdEIP

TkwdEIP

)/exp(

edMPd

LBedMPu

weISP

TkweISP

)/(

)/exp()/(

A. Bravaix et al., IRPS, 2009 C. Guerin et al., APL, 2009

)/( eISdEI dMPd

56


MP-process

Si-H → truncated harmonic oscillator

System of rate equations:

Finally: AI → empirical factor

The simplified solution for the MP-process:

Square root time dependence

edMPd

LBedMPu

weISP

TkweISP

)/(

)/exp()/( 2

1

11

010

)()(

-

MPpassNemiNdNu

N

iiuiidi

ud

NPnnPnPdt

dn

nnPnnPdt

dn

nPnPdt

dn

lll

l

tPPNtN lN

duemiMP /)( 0

57


Low Id regime High carrier energies

“Hot-carrier” regime

SP-mechanism plays the dominant role:

High electron flux

Low carrier energies

MP-process dominates

Intermediate case

Moderate Id and Vds

Governed by electron-electron scattering:

Real device stress/operation conditions → all the modes are present

KMP, KEES, KSP – prefactors

/2/1

/2/1

)]/([

)/exp()]/()[(~/1

B

B

E

dds

LBemi

E

sdsMP

WIV

TkEWIV

mdbdSPSPSP IIWICR //~/1

mdbdEESEESEES IIWICR //~/12

MPMPEESEESSPSPd KKK ///~/1

C. Guerin et al., IRPS, 2007

58


The fitting scheme for the SP-process rate

Used to better fit the CP data

Knee energy

From Rauch and LaRosa paradigm

No carrier transport sub-task

DF is substituted by empirical factors

Nit: agreement is good

eVEES

eVEES

eVEconstS

eVES

SP

SP

SP

SP

5.2,)5.1(

5.25.1),3exp(

9.15.1,

5.1,0

11

R.M. Randrimihaja et al., Microel. Reliab., 2012

59


Advantages

The microscopic and device levels are connected

Shortcomings

Simplified treatment of carrier transport

DF→ empirical factors

Electron-electron scattering cannot be treated as a separated mode because

It affects the carrier DF

Defines interplay between SP- and MP-processes

Final parameterization is

Based on fitting parameters

Not on physical mechanisms

60

Outline


Main peculiarities




Carrier transport




Conclusions

61

Our Model: Structure

The model

Features of previous approaches

Linking all the levels related to this effect

A physics-based model contains

Carrier transport module

Module describing the defect build-up

Module for simulation degraded devices

Carrier transport

Full-band Monte-Carlo simulator MONJU

Allows to thoroughly evaluate the DF

For a particular device architecture

Ch. Jungemann, B. Meinerzhagen, Hierarchical Device Simulation, Springer Verlag, 2003

MiniMOS-NT, Device and Circuit Simulator, Institute for Microelectronics, TU Wien

GTS Framework, Global TCAD Solutions, Vienna, Austria

62

Our Model: Comprehensive?

Considers only the electron contribution Calibrated for 5V n-MOSFET (0.35 μm node), Lch = 0.5 μm

Successfully represented Idlin degradation

Further verification

Should properly describe HCD for different channel lengths

We used identical device architecture

Differing in channel lengths: Lch = 0.5, 1.2 and 2.0 μm

Stress conditions: Vgs = 2.0V

Vds = 6.25V

T = 250C

Calibration: to represent the Idlin degradation

S. Tyaginov et al., IPFA, 2010

S. Tyaginov et al., SISPAD, 2011

63

Our Model: Comprehensive?

The first version of our model fails

While trying to capture Idlin degradation in different MOSFETs

ΔIdlin = (Idlin0 – Idlin(t))/Idlin0

Lch = 1.2, 2.0 μm:

Theoretical ΔIdlin is less than experimental

Nit by electrons: peaks outside the channel

Average interface trap density: <Nit>

Stronger degradation ↔ less <Nit>

Longer devices: less sensitive to electron-induced Nit

64

Our Model: Missing Contribution

Mechanism generating interface states closer to the channel

Secondary generated holes

By impact ionization

Accelerated by the electric field

Interface states shifted to the source

The same Nit stronger affecting the device performance

Holes should be considered

S. Tyaginov et al., SISPAD, 2011

65

Our Model: Electrons and Holes

Superposition of electron and hole AIs

Monte-Carlo simulator → DFs for electrons and holes

DFs → the carrier acceleration integral

The same functional structure for:

SP- and MP-processes and for electrons and holes

SP-process

First order kinetics

MP-process

Truncated harmonic oscillator

tII

SPhSPhSPeSPeSPeNtN ,,,,1)( 0

2/1

0 1)(

t

N

d

u

pass

emiMP

emi

l

eP

P

PNtN

thE

dEEEvEgEfI )()()()(

ehMPeMPd

LBehMPeMPu

wIIP

TkwIIP

,,

,, )/exp(

66

Our Model: Verification and Results

Secondary holes

Generated by impact ionization

Due to the hot-electron injection

Holes are accelerated by the electric field

Towards the source

The hole AI is considerably shifted

towards the source

67


Model represents degradation

For different channel lengths

With the same set of parameters

For Lch = 0.5:

Hole contribution may be neglected

68

Outline


Main peculiarities




Carrier transport




Conclusions

69


The driving force of HCD

Nit(x) profiles feature two peaks

In good agreement with the results of our HCD model

Peaks: by primary channel electrons and secondary generated holes

Correspond to the maxima of electron and hole acceleration integrals

S. Tyaginov et al., SISPAD, 2011 I. Starkov et al., IRPS, 2012

70

Open Issues: Pro Not

Do bulk oxide traps contribute?

Arguments pro: Threshold voltage turn-around effect

Hot-carrier stress at WCC

The threshold voltage Vth was monitored up to 105s

First Vth decreases

Due to h+ trapping in the oxide bulk

After 10ks Vth increases

Due to trapping of e- by interface traps

At ~100 ks: compensation

Supported by Nit(x) and Not(x) profiles

Extracted from CP data

I. Starkov et al., IRPS, 2012

71


Charge-pumping signal: no saturation plateau

Only interface traps

ICP vs. Vgh (varying high-level technique)

saturates, thereby demonstrating a plateau

This tendency is not pronounced

CP current continues to increase

Contribution of bulk oxide traps

72


Turn-around effect of Idlin

p-DEMOS, 0.35 design

Stressed at Vds = -23V, Vgs = -12V

Short stress times:

Electrons stored in bulk oxide traps

In the Lp region

|Idlin| increases

Long stress times:

Holes trapped by interface states

In the Lov region

|Idlin| decreases

Result: turn-around effect

J.F. Chen et al., Jpn. JAP, 2009

73

Open Issues: Contra Not

HCD demonstrates no (or weak) recovery

Switching oxide traps in BTI

Responsible for BTI recovery

By analogy

Oxide traps in HCD → Recovery should be pronounced

Only high-voltage devices (LDMOS, DEMOS) demonstrate recovery

Scaled CMOS transistors: No recovery

Contradictions → Reconciliation in further research

T. Grasser et al., TED, 2011 H. Park et al., IIRW, 2008 R. O’Connor et al., IRPS, 2008

74

Conclusions part I: Experimental Facts

Due to generation of defects at or near the interface

Strongly localized

Temperature behavior

Long-channel devices: HCD is suppressed at elevated temperatures

Ultra-scaled devices: HCD becomes more severe at elevated temperatures

Contribution of bulk oxide traps

Results in turn-around effects

Charge-pumping current does not demonstrate a plateau

75

Conclusions Part II: Complexity of HCD

Carrier transport

Scattering mechanisms

Carrier energy distribution: “cold” vs. “hot” carriers


Dissociation of Si-H bonds

Single- and multiple-carrier processes

Contributions of electrons and holes

Device level

Worst-case conditions of HCD

Their change with MOSFET downscaling

Degradation of device characteristics with time

76

References

A. Abramo et al., IEDM Techn. Dig., p. 301, 1995.

M. Ancona et al., IEEE Trans. Elecron. Dev., v. 35, No. 12, p. 2221, 1988.

A. Bravaix et al., Proc. 47th Intern. Reliab. Phys. Symp. – IRPS, p. 531, 2009.

R. Biswas et al., Appl. Phys. Lett., v. 72, No. 26, p. 3500, 1998.

J. Bude, VLSI Symp. Techn. Dig., p. 101, 1995.

J.F. Chen et al., Jpn. Journ. Appl. Phys., v. 48, pap. No. 04C039, 2009.

P.A. Childs and C.C.C. Leung, Journ. Appl. Phys., v. 70, No. 1, p. 222, 1996.

D.J. DiMaria and S.W. Stasiak, Journ. Appl. Phys., v. 65, No. 6, p. 2342, 1989.

D.J. DiMaria and J.H. Stathis. Journ. Appl. Phys., v. 89, No. 9, p. 5015, 2001.

M. Fischetti and S. Laux, IEDM Techn. Dig., p. 305, 1995.

A. Ghetti et al., IEDM Techn. Dig., p. 893, 1998.

T. Grasser et al., IEEE Trans. Elecron. Dev., v. 58, No. 11, p. 3652, 2011.

GTS Framework, Global TCAD Solutions, Vienna, Austria.

C. Guerin et al., Proc. 45th Intern. Reliab. Phys. Symp. – IRPS, p. 692, 2007.

C. Guerin et al., Journ. Appl. Phys., v. 105, No. 11, pap. No. 114513, 2009.

C.R. Helms and H.E. Poindexter, Rep. Prog. Phys, v. 57, No. 8, p. 791, 1994.

K. Hess et al., Physica E, v. 3, p. 1, 1998.

K. Hess et al., Circ. Dev. Mag., p. 33–38, 2001.

T. Hori et al., IEEE Trans. Elecron. Dev., v. 39, No. 10, p. 2312, 1992.

F.-C. Hsu and K.-Y. Chu, IEEE Elecron. Dev. Lett., v. 5, No. 5, p. 148, 1984.

C. Hu et al., IEEE Trans. Elecron. Dev., v. 32, No. 2, p. 375, 1985.

T.-Y. Huang et al., IEDM Techn. Dig., p. 742, 1986.

Ch. Jungemann, B. Meinerzhagen, Hierarchical Device Simulation, Springer Verlag, 2003.

77

References

J.W. Lyding et al., Appl. Phys. Lett., v. 68, No. 18, p. 221, 2526.

J.W. Lyding et al., Appl. Surf. Sci., v. 130-132, p. 221, 1998.

H. Kufluoglu and M. Alam, Journ. Comput. Electron., v. 3, No. 3-4, p. 165, 2004.

W. McMahon et al., 2002 Int. Conf. Mod. Sim. Micro, v. 1, p 576, 2002.

W. McMahon, et al., IEEE Trans. Nanotech., v. 2, No. 1, p. 33-38, 2003.

MiniMOS-NT, Device and Circuit Simulator, Institute for Microelectronics, TU Wien.

T. Mizuno et al., IEDM Techn. Dig., p. 695, 1992.

R. O’Connor et al., Proc. 46th Intern. Reliab. Phys. Symp. – IRPS, p. 324, 2008.

H. Park et al., Fin. Rep. Intern. Integrated Reliab. Workshop – IIRW, p. 44, 2008.

P. Penzin et al., IEEE Trans. Electron Dev., v. 50, No. 6, p. 1445, 2003.

B. Persson and P. Avouris, Surface Science 390, No. 1-3, p. 45, 1997.

R.M. Randrimihaja et al., Microel. Reliab., in press, 2012.

S.E. Rauch et al., IEEE Elecron. Dev. Lett., v. 19, No. 12, p. 463, 1998.

S.E. Rauch and G. LaRosa, IEEE Trans. Dev. Material. Reliab., v. 1, No. 2, p. 113, 2001.

S.E. Rauch and G. LaRosa, Proc. 43rd Intern. Reliab. Phys. Symp. – IRPS, p. 708, 2005.

B. Ricco et al., IEDM Techn. Dig., p. 84, 1992.

I. Starkov et al., in Proc. of 50th Intern. Reliab. Phys. Symp. – IRPS, p. XT.7.1, 2012

K. Stokbro et al., Phys. Rev. Lett., v. 80, No. 12, p. 2618, 1998.

B. Tuttle and Ch.G. Van de Walle, Phys. Rev. B, v. 59, No. 20, p. 12884, 1999.

S. Tyaginov et al., Proc. Int. Symp. Phys. Fail. Analys. Integr. Circ. – IPFA, p. 341, 2010.

S. Tyaginov et al., Proc. Int. Conf. Simulat. Semicon. Process. Dev. – SISPAD, p. 127, 2011.

D. Varghese et al., IEEE Electon Dev. Lett., v. 26, No. 8, p. 572, 2005.

Modeling of Hot-carrier Degradation: Physics and ... · Employment of LDD structures V ds is scaled...

Documents

Transcript of Modeling of Hot-carrier Degradation: Physics and ... · Employment of LDD structures V ds is scaled...