Kinetics of Silicon Oxidation in a Rapid Thermal Processor 2/Day-1/Asad_NCP_March_201… · Asad M. Haider ; March 01, 2010 ; NCP Kinetics of Silicon Oxidation in a Rapid Thermal

Asad M. Haider ; March 01, 2010 ; NCP

Kinetics of Silicon Oxidation in aRapid Thermal Processor

Asad M. Haider, Ph.D.

Texas InstrumentsDallas, Texas

USA

Presentation at the National Center of PhysicsInternational Spring Week 2010

IslamabadPakistan

March 01, 2010


PRESENTATION OUTLINE

• Introduction and motivation to study Si oxidation

• Mechanism of Si oxidation

• Mathematical model for Si oxidation

• Hardware design of a Rapid Thermal Processor (RTP)

• Experimental data and the model parameter estimation

• Oxide quality

• Conclusions


INTRODUCTION: Importance of SiO2 in SC Industry

During semiconductor device manufacturing SiO2 is thermally grown to be used as a:

a) Gate oxideb) Isolation oxide liner between devices (STI liner)c) Masking element (for eg., during ion implantation)d) Surface passivation (for eg., Pad oxide. A sacrificial layer for contamination control)

Gate

Gate Oxide

Source Drain

Substrate Isol

atio

n

Isol

atio

n Lg

Please note the difference between “grown” SiO2 and “deposited” SiO2

This presentation is about thermally “grown” SiO2


1. To indirectly measure across wafer temperature uniformity of a Rapid Thermal Processor in > 900C range.

2. Understand the Si oxidation kinetics in a RTP chamber and measure Deal-Grove oxidation model parameters for < 30nm thick oxides.

3. Understand the impact of various process parameters on SiO2 growth in a RTP chamber.

4. Compare the oxide quality grown in a RTP with that grown in a furnace.

MOTIVATION TO STUDY OXIDATION IN RTP


OXIDATION OF SILICON

Si has great affinity for oxygen and is easily oxidized in a number of ways:

1. Chemical oxidationBoil Si in HNO3,,for example.

2. Anodic oxidationIn an electrolytic bath use Si as an anode and a noble metal as a cathode.

3. Plasma oxidationUses ions of an oxidant species to grow oxide film.

4. Thermal oxidation• Used exclusively in semiconductor device fabrication.• Gives by far the best quality oxide.• Typically done in a furnace.• Two types of thermal oxides:

Si + O2 SiO2 Dry oxidationSi + 2 H2O SiO2 + 2 H2 Wet Oxidation

• Dry oxidation: Slow, high density, good quality Thin gate oxides• Wet oxidation: Fast, low density, poorer quality Thick mask/passivation

This study looks at kinetics of dry oxidation in a Rapid Thermal Processor.


MECHANISM OF Si OXIDATION

Question: Is it the Si atoms that diffuse through the oxide to react with O2 at the oxide surface or is it the O2 that diffuses through the oxide to react with Si at the Si/SiO2 interface?

Answer: For thermal oxidation, it has been established through radioactive tracer studies that it is the O2 that diffuses through the oxide and reacts with Si at the Si/SiO2 interface.

O2

SiSiO2

Consequently, thermal oxidation always takes place on fresh Si surface rather than the original surface that may have been exposed to ambient contaminants.

Next, we look at a detailed mathematical model for the oxidation of Si.


MATHEMATICAL MODEL FOR SILICON OXIDATION

Si + O2 SiO2 Dry Oxidation

Gas Oxide Silicon

Cg

Cs

Co

Ci

δ

N1 N2 N3

x

Cg ≡ Concentration of oxidant molecules in the bulk gasCs ≡ Concentration of oxidant molecules immediately adjacent to the oxide surfaceCo ≡ Equilibrium concentration of oxidant molecules at the oxide surfaceCi ≡ Concentration of oxidant molecules at the Si/SiO2 interface

Note: i) Cg > Cs due to depletion of the oxidant at the oxide surfaceii) Cs > Co due to solubility limits of the oxide

δ ≡ Oxide thickness at a given timeNi ≡ Flux of oxidant molecules

Deal and Grove, J. Appl. Physics, vol 36, p 3770, (1965)


N1 = Oxidant flux from bulk gas to the oxide surface

( )sgm CCkN −=1 (Eq. 1)

N2 = Oxidant flux through the oxide

dxdCDvC

dxdCDN −=+−=

r2

Integration across the oxide film gives:

( )δ

io CCDN −=2 (Eq. 2)

ikCN =3(Eq. 3)

Henry’s law dictates that:so HpC = and gHpC =*

Gas Oxide Silicon

Cg

Cs

Co

Ci

δ

N1 N2 N3

x

pg

ps

C*

C* = Equilibrium conc in bulk oxide

Therefore, Eq. 1 becomes:

( )om CC

HRTkN −= *

1 (Eq. 4)

MATHEMATICAL MODEL FOR SILICON OXIDATION – Contd.



⎟⎟⎠

⎞⎜⎜⎝

⎛++

⎟⎠⎞

⎜⎝⎛ +

=

Dk

kkHRT

Dk

CC

m

o δ

δ

1

1*

⎟⎟⎠

⎞⎜⎜⎝

⎛++

=

Dk

kkHRT

CC

m

i δ1

1*

Gas Oxide Silicon

Cg

Cs

Co

Ci

δ

N1 N2 N3

x

pg

ps

C*

C* = Equilibrium conc in bulk oxide

Express Co and Ci in terms of measurable quantities.At steady state: N1 = N2 = N3This results in:

(Eq. 5)

(Eq. 6)

Case 1: Mass transfer controlled process:Oxide growth rate depends only on how fast oxidant is supplied to the Si/SiO2 interface.Hence, D << k ⇒ Ci ~ 0 and Co ~ C*

Case 2: Kinetics controlled process:Oxide growth rate depends only on how fast the oxidant reacts at the Si/SiO2 interface.

Hence, D >> k ⇒

⎟⎟⎠

⎞⎜⎜⎝

⎛+

==

m

oi

kkHRTCCC

1

*


Oxide Growth Rate:Let Γ be the number of oxidant molecules per unit volume of the oxide film. Then,

( )⎟⎟⎠

⎞⎜⎜⎝

⎛++

===Γ

Dk

kkHRT

kCkCNdtd

m

i δδ

1

*

3 (Eq. 7)

Integrating Eq. 7 with initial condition: At t = 0 ; δ = δi results in:


( )EtBA +=+ δδ 2 (Eq. 8)

Where,

⎟⎟⎠

⎞⎜⎜⎝

⎛+=

mkHRT

kDA 12

Γ=

*2DCB

BAE ii δδ +

=2

A and B are the only two model parameters to be found experimentally.



Special Cases:

A. For very short times, δ is very small and the process is kinetics limited. In this regime Eq. 8 becomes:

( )EtAB

+=δ (Eq. 9)

B. For very long times, δ is pretty thick and the process is diffusion limited. In this regime Eq. 8 becomes:

Bt=δ (Eq. 10)

To find model parameters A and B requires collecting oxide thickness vs. time data.

Since A and B are in turn functions of temperature, oxide thickness data needs to be collected at different temperatures in order to develop a general equation to predict oxide thickness as a function of time and temperature, δ = δ(time, temperature)

But first, let us look at the hardware design of a RTP chamber.


DESIGN OF RTP (Rapid Thermal Processor)

Transfer chamber and chambers A and B.


View of an open RTP chamber.

Close-up of the reflector plate. Pyrometers and lift pin holes visible.

DESIGN OF RTP – Contd.


Wafer Edge Ring and Support Cylinder Assembly Schematic


Assembled parts: SiC wafer edge ring sitting on top of the support cylinder around the reflector plate.




Reflector plate showing raised wafer lift pins.


RTP Centura Lamp Zones and Temperature Probe Locations



Close up of the RTP multi-zone lamp heater assembly capable of precision controlled temperature ramp rates of >100C/s.


Oxide Growth Rate vs. Pressure at 1050C

0

0.2

0.4

0.6

0.8

1

1.2

0 100 200 300 400 500 600 700 800 900

Pressure, torr

Oxi

de G

row

th R

ate,

A/s

KEY PROCESS PARAMETERS AND THEIREFFECT ON OXIDATION KINETICS

Γ=

*2DCBRecall, ; As P increases, C* increases


All Si oxidation tests were conducted in O2 ambient at 5 slm at a chamber pressure of 780 torrat various temperatures.

SiO2 Growth Rate Vs. O2 Flow RateT = 1050C ; P = 780 torr

0

0.2

0.4

0.6

0.8

1

1.2

0 1 2 3 4 5 6 7 8

O2 Flow Rate, slm

SiO

2 G

row

th R

ate,

A/s


Next, estimate model constants A and B by doing a least squares fit of the model to the experimentally collected Si oxidation data.

Arrhenius Plot for SiO2 GrowthP = 780 torr, O2 = 5 slm

y = -13808x + 10.493R2 = 0.9982

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.00072 0.00073 0.00074 0.00075 0.00076 0.00077 0.00078 0.00079 0.0008 0.00081 0.00082 0.00083

1/Temp, 1/K

ln (R

ate)

, A/s

E = 114.8 kJ/mole


EXPERIMENTAL DATA AND CALCULATIONOF MODEL PARAMETERS

Oxide Growth at 1050C in RTP Reactor

0

20

40

60

80

100

120

140

160

180

200

0 50 100 150 200 250 300 350

Oxidation Time, s

Oxi

de T

hick

ness

, A

Theory Measured

A = 40 °AB = 131 °A2/s


Oxide Grow th at 1075C in RTP Reactor

0

50

100

150

200

250

0 50 100 150 200 250 300 350

Oxidation Time, s

Oxi

de T

hick

ness

, A

Theory Measured

EXPERIMENTAL DATA AND CALCULATIONOF MODEL PARAMETERS

A = 55 °AB = 188.5 °A2/s

Repeat these tests at multiple temperatures to get dependency of A and B on temperature.


Rate Constant A vs. Temperature

y = 0.4582x - 439.45R2 = 0.9971

0

10

20

30

40

50

60

70

950 975 1000 1025 1050 1075 1100 1125

Temp, C

A, A

ng


Parabolic Rate Constant B vs. Temperature

y = 0.0145x2 - 28.029x + 13567R2 = 0.9975

0

50

100

150

200

250

300

350

950 975 1000 1025 1050 1075 1100 1125

Tem p, C

B, A

ng^2

/s


( )EtBA +=+ δδ 2

45.4394582.0 −= TA

13567029.280145.0 2 +−= TTB

BAE ii δδ +

=2

δ = Oxide thickness grown at any timeδi = Initial oxide thicknesst = TimeT = Temperature

Semi Empirical Model to Predict SiO2 Thickness NearAtmospheric Pressures in RTP


OXIDE QUALITY

SiO2 SiOx Si-----

+++++

+++++

-----

Na+

K+

Mobile ionic charges Oxide trapped charges Fixed oxide charges Interface trapped charges

Source:• Mostly humans.• Contaminated water

or if it is not fully deionized.

Effect:Wreaks havoc on transistor characteristics.

Fix:Use chlorine oxidation – bubble O2 through TCE. Be careful, too much Cl will result in “halogen pitting”.

Source:• Mechanical damage in Si

wafer.• Dangling Si bonds left un-

reacted after oxidation.

Effect:Trap and de-trap electrons affecting MOS device performance.

Fix:Do a low temp (~450C) anneal in H2 ambient post oxidation.

Source:• Incomplete oxidation

Effect:Pushes VT in –ve direction

Fix:At the end of oxidation step purge the system with N2 or Ar gas and then drop the temperature.

Source:• Exposure to radiation

environment.• Hot electron effect in

short channel MOSFET devices.

Effect:Interferes with electronic activity.

Fix:Typically not caused by processing itself.

Best way to tell the quality of oxide is by measuring the charges in it.


SUMMARY AND CONCLUSIONS

• Deal and Grove model was successfully applied to oxidation of Si in a RTP reactor for oxide thicknesses less than 30nm.

• Model parameters A and B were empirically found as a function of temperature at 780 torr to obtain a generalized model capable of accurately predicting dry Si oxidation rates between 975C and 1100C for oxide thicknesses less than 30nm.

• Activation energy of dry Si oxidation in RTP was found to be 115 kJ/mole.

• Discussed various types of charges in SiO2 that determine the quality of oxide and how to mitigate them.

Documents

Kinetics of Silicon Oxidation in a Rapid Thermal Processor 2/Day-1/Asad_NCP_March_201… · Asad M. Haider ; March 01, 2010 ; NCP Kinetics of Silicon Oxidation in a Rapid Thermal