OXIDATION- Overview

Process Types Details of Thermal Oxidation Models Relevant Issues

Uses

As a part of a structure e.g. Gate Oxide

For hard masks e.g. In Nitride Etch, implant mask ...

Protecting the silicon surface (Passivation ) Insulator (ILD/IMD) As part of ‘mild etch’ (oxidation / removal cycles)

Whether useful or not, automatically forms in ambient Native Oxide ( ~ 20 A thick) except H-terminated Si (111)

Processes

Thermal Oxidation (Heating) Dry vs Wet

Electrochemical Oxidation (Anodization)

Oxide (and nitride) adhere well to the silicon good insulator Breakdown voltage 10 MV/cm

==> Can make a very thin gate

Structure Tetrahedral Structure

each Si to four O each O to two Si

Single crystal quartz (density 2.6 g/cm3) Fused silica (density 2.2 g/cm3)

Reaction with water

©Time Domain CVD

2 0Si O Si H Si OH Si OH

Si-OH termination is stable structure is more porous than Si-O-Si

Thermal Oxidation

Dry oxidation

2 2Si O SiO 2 2 22 2Si H O SiO H

Dense oxide formed (good quality, low diffusion) slow growth rate

NEED TO KEEP WATER OUT OF THE SYSTEM

Wet oxidation

Overall reaction Relatively porous oxide formed (lower quality, species diffuse faster)

Still good quality compared to electrochem oxidation, for example

faster growth rate

Wet oxide for maskingDry oxide for gate ox

Wet Oxidation Proposed Mechanism

2Si O Si H O SiOH SiOH

22 2Si OH Si Si Si O Si H

Hydration near Silicon/ Silicon oxide interface

Oxidation of silicon

Hydrogen rapidly diffuses out

Some hydrogen may form hydroxyl group

21

2Si O H SiOH

Diffusivities in Oxide

Oxygen diffuses faster (compared to water) Sodium and Hydrogen diffuse very fast

Water

Oxygen

Hydrogen

Sodium

1/T

Dif

fusi

vity

(lo

g sc

ale)

Oxide Growth (Thermal)

SiOxide

Original Si surface

To obtain 1 unit of oxide, almost half unit of silicon is consumed (0.44) Oxidation occurs at the Si/SiO2 interface i.e. Oxidizing species has to diffuse through ‘already existing’ silicon oxide

Oxide Growth (Thermal)

SiliconOxideAir (BL) At any point of time, amount of oxide is variable ‘x’ Usually, concentration of oxidizing species (H2O or O2) is sufficiently high in gas phase==> Saturated in the oxide interface x

Distance

Con

cent

rati

on o

i

Oxidation Kinetics At steady state

diffusion through oxide = reaction rate at the Si/SiO2 interface

Oxygen diffuses faster than Water However, water solubility is very high (1000 times) ==> Effectively water concentration at the interface is higher ==> wet oxidation fasterdN

J Ddx

( )o iN N

Dx

iRate k N oi

NN D

kx D

At steady state

Diffusion

Reaction

Oxidation Kinetics

Oxide Growth Rate

oDNJ

Dx k

Flux atsteady state

dx

dt

= Flux/ # oxidizing species per unit volume (of SiO2) n = 2.2 × 1022 cm-3 for O2

= 4.4 × 1022 cm-3 for H2O

J

n

oDNdxDdt x k

0ix x at t

EqnInitial Condition

6.023x1023 molecules =1 mol of oxide = x g of oxide = y cm3 of oxide (from density) 2.2 x 1022 molecules/cm3

One O2 per SiO2

Two H20 per SiO2

Deal-Grove Model

2 022( )

DNDx x t

k n

Solution

2i ix x

BBA

2DA

k

2 oDNB

n

where

2x x

tBB

A

OR

is the time needed to grow the ‘initial’ oxide

A and B depend on diffusivity “D”, solubility and # oxidizing species per unit volume “n” A and B will be different for Dry and Wet oxidation

Bruce Deal & Andy Grove

Linear & Parabolic Regimes

Linear vs Parabolic Regimes Kinetic Controlled vs Mass Transfer Controlled Initially faster growth rate, then slower growth rate

( )B

x tA

Very short Time

2 ( )x B t

Longer Time

If one starts with thin oxide (or bare silicon)

12

2

40.5 1 ( ) 1

Bx A t

A

2

( )4

At

B

2

4

At

B

2

4

At

B

Exponential Regime

Hypothesis 1 Charged species forms holes diffuse faster / set up electrical field diffusion + drift ==> effective diffusivity high space charge regime controls length = 15 nm for oxygen, 0.5 nm for water ==> wet oxidation not affected

For dry oxidation, one finds that is not zero in the model fit A corresponding to an initial thickness of 25 nm provides good fit Initial growth at very high rate Approximated by exponential curve

If one starts with bare oxide

Exponential Regime Hypothesis 2

In dry oxidation, many ‘open’ areas exist oxygen diffuses fast in silicon hence more initial growth rate once covered by silicon di oxide, slow diffusion

Hypothesis 3 Even before reaction (at high temp), oxygen dissolved in silicon (reasonable diffusion) once temp is increased, 5 nm quick oxide formation

Temp Variation of Linear/Parabolic Coeff

Linear [B/A] Parabolic [B]Solubility and Diffusion function of temp

© May & Sze

Effect of Doping Doping increases oxidation rate Segregation

ratio of dopant in silicon / dopant in oxide

e.g. Boron incorporated in oxide; more porous oxide more diffusion parabolic rate constant is higher

P not incorporated in oxide no significant change in parabolic rate constant

© May & Sze

Issues Na diffuses fast in oxide Use Cl during oxidation

helps trap Na helps create volatile compounds of heavy metals (contaminant from furnace etc) use 3% HCl or Tri chloro ethylene (TCE)

Ref: VLSI Fabrication Principles by S.K. Ghandhi

Electrochemical

Use neutral solution and apply potential Pt as counter electrode (Hydrogen evolution) Use Ammonium hydrogen Phosphate or Phosphoric acid or ammonia solution Silicon diffuses out and forms oxide Increase in oxide thickness ==> increase in potential needed

self limiting Oxide quality poor Used to oxidize controlled amount and strip

for diagnosis