Lecture 2; Thermal Oxidation

SiO2 in micro-nano fabrication a) gate in MOS structures b) device isolation- field oxide c) electrical isolation in multilevel metallization systems (ILD- Inter Layer/level Dielectric) d) surface /protection e) implant mask f) diffusion mask g) etch mask (lithography, MEMS, etc) i) Flash memory tunnel oxide j) DRAM capacitor oxide Several techniques to grow thin dioxide layers a)Thermal oxidation, etc (wet, dry, LTO, TEOS, TCA) b) Chemical Vapour Deposition ( CVD, PECVD, ALD) c) Sputter deposition d) Physical/Electronbeam Evaporation FHR sputter

LP

LP TEOS Features of SiO2 (thermal, etc):

SiO2 can be etched with HF, which leaves Si unaffected SiO2 is a good diffusion barrier for B, P, As SiO2 is a good electric insulator, refraction index=1.462 SiO2 has high dielectric breakdown field, 500 V/μm SiO2 growth on Si→ clean Si/SiO2 interface

Centrotherm

Fused silica Horisontal pipe furnaces at MC2 Dry oxidation process (800-900 C);

Si(solid)+O2(gas)→ SiO2

Wet oxidation process (900-1200);

Si(solid)+ O2(gas) +2H2O(gas)→ SiO2+2H2

2H2+O2 ---2H2O

Nitride Nitride Nitride Nitride Nitride Nitride mask mask mask mask mask mask 44% Si

Time; t=0 t1>0 t2>t1

Clean and defect free interface!

Fig 15.6- LOCOS process Oxide growth model The Deal-Grove model (1965)

Growth process limited by

P(O2)=pg J= molecule flux (J1, J2, J3)

Cg= oxygen concentration in gas “far away” from surface”, pg Cs=concentration in gas on the surface Co= concentration of O2 in SiO2 at the surface

Ci= concentration of O2 at the SiO2-Si interface The Deal-Grove model

Growth process limited by;

P(O2)=pg J= oxygen molecule flux

Transport of O2 to SiO2 surface across “Stagnant gas layer”; J1 Diffusion of O2 through SiO2; J2 Chemical reaction rate at interface; J3 The first flux; J1; is the diffusion of molecular oxygen through the “stagnant layer” (Ficks 1st law) C  C J  D g s Note; convection and other migration mechanism 1 O2 t s l through the layer is neglected! tsl is thickness of stagnant layer

This flux is approximated to;

(hg is the mass transport coefficient)

J1=hg(Cg-Cs) The second flux; J2; is the diffusion of molecular Co Ci oxygen across the growing oxide film, J 2  DO2 tox tox is the thickness of oxide layer The third flux; J3 is the flow of Oxygen molecules reacting with Silicon to form SiO2 as determined by chemical reaction kinetics

J3=ksCi In equilibrium; J1=J2=J3=J

And adding the Henry’s law (1803)

Co=Hps

(concentration of an adsorbed species at the surface of a solid is proportional to the partial pressure of that species in the gas just above the solid, H- Henry’s gas constant) n P and using ideal gas law; C   s s V kT

Co=Hps=HkTCs O2 concentration at interface can be calculated; Ci

J2 J J 3 C  HP  HkTC 1 C C 0 s s J  h (C C )  D o i  k C g g s O2 s i + n Pg tox C   g V kT

???? O2 concentration at interface can be calculated; Ci

J 2 J3 J1 C0  HPs  HkTCs Co Ci J  hg (C g Cs )  DO2  k s Ci + n Pg tox C   g V kT

J HPg / ks Eq. 4.9 3 Ci   ks 1 tox 1      h D ks 

Mass transfer Reaction rate Diffusion in SiO2 The oxidation rate; R

R (m/s); tox= oxide thickness

J J3 ksCi dtox Hk s Pg R      Eq. 4.10 N1 N1 N1 dt  ks kstox  N1 1    h D 

R; divide total number of O2 flux at interface(=J3) by

N1= the number of oxygen molecules per unit volume of SiO2 22 3 (SiO2 Atomic Density; 2.27x10 molecules/cm ) Assuming that at time 0 the oxide thickness is to and integrating tox t  k k t  Hk s Pg 1 s  s ox dt  dt  h D ox  N to   0 1

the solution of this differential equation is Assuming that at time 0 the oxide thickness is to and integrating

tox t  k k t  Hk s Pg 1 s  s ox dt  dt  h D ox  N to   0 1

the solution of this differential equation is

2 Eq. 4.11 t ox  Atox  B(t )

2  1 1  2DHPg t o  At Eq. 4.12 A  2D   B    o  k h   s  N1 B The Linear and Parabolic Rate coefficients

The rate equation has two important limiting forms. for sufficiently thin , one can neglect the quadratic term

2 B t ox  Atox  B(t ) → t  (t  ) ox A

B/A is called the linear rate coefficient if the oxide is sufficently thick, the linear term can be neglected

2 2 t ox  Atox  B(t ) → t ox  B(t )

B is called the parabolic rate coefficient 2 t ox  Atox  B(t )

tox tox Dry oxide ; 800-1200 C, 0,1 micron/h is denser and used for gate oxide

Wet oxide ; 750-1100 C; 1 micron/hr is more porous and poorer diffusion barrier Used for etch mask oxide, field oxide much higher oxidation rate due to higher solubility of H2O (lover diffusivity!) • Example 4.1 Calculate the amount of grown during 120 min, at 920 C and using steam (wet) oxidation. Assume that the wafer initially had 100 nm oxide 2 t ox  Atox  B(t ) Eq 4.11

A=0.50 and B=0.203 and t= 2 According to eq. 4.12 = (0.1x0.1 + 0.5x0.1)/0.203 = 0.295 h

2 And via t ox  Atox  B(t )

2 x tox= (-0.5+sqrt(0,5 +4 0.203(2+0.295)))/2= 0.48 mm

Thin film approximation; tox=0,93 Thick film approximation; tox=0,465 Rate coefficients

Linear Parabolic Oxides thicker than 1 μm requires very long annealing times, which can influence dopant redistribution in wafer. (also, Si is consumed!)

Increasing the Cg ( the equilibrium concentration of oxygen in the gas/ partial pressure of oxygen in gas) can speed up the oxidation and minimize dopant diffusion 4.3 The initial oxidation regime

The Deal-Grove model fits the oxidation rate over a broad range of parameters, if the linear and parabolic parameters are known

2 t ox  Atox  B(t )

lim dt B ox  t  0 dt A

Oxidation rate is enhanced in the initial oxidation stage!! Different models for explaining the enhanced oxidation rate based on different mechanisms.

- a) enhancement of arrival rate at interface, the electric field, indicating diffusion by O2 . b) “worm holes”, microchannels in oxide of 50Å exist!? c) stress due to different thermal expansion coefficients d) increase in oxygen solubility in the oxide. e) the oxidation process occurs over some finite thickness, ie no abrupt interface.

J3 HPg / ks Ci   ks 1 tox 1      h D ks  C k  s C L

Redistribution of dopants during oxidation

Nearly all substrates has some doping content. This doping concentration profile will change during oxidation, due to the segregation effect; C (Si) m  i Ci (SiO2 )

Ci= concentration of impurity at equilibrium

m>1; impurity accumulation in Si

m<1; impurity accumulation in SiO2 C k  s C L

Impurity element concentration profile depend on a) M (solubility) b) Impurity Diffusivity in Si Boron Boron c) Impurity Diffusivity in SiO2 d) Oxide growth rate

4 classes of behaviour; Fig 4.16 m<1 P, As Ga x x Fig. A; D (SiO2)D (Si)

m>1 x x Fig. C; D (SiO2)D (Si) (e.g. Ga) The common silicon dopants enhance the oxidation rate at high concentrations

B-Boron P-Phosphorous Characterisation of thin SiO2 films Optical properties (refraction index, n and k) Electrical properties Structural defects, impurities, etc

Woolam Ellipsometer (thickness, 3D mapping, etc)

Sagax ellipsometer SUMMARY-Thermal Oxidation

Deal-Grove oxidation model Thin (linear) vs. Thick (parabolic) region Rate depends on oxidation conditions Implantation

http://www.casetechnology.com/links.html The thermal diffusion process for introducing dopants in the substrates has limitations a) Can not exceed solid solubility of dopant b) Difficult to achieve light doping (below surface solubility) for shallow profiles c) Difficult to control depth distribution (time, D)

Ion implantation is used since 1973 and preferred because a) Controlled, low or high dose can be introduced b) Depth of implant can be controlled c) Less lateral distribution than thermal diffusion (below mask)

Typical ion energies; 10-300 keV Typical doses: 1011 to 1016 at/cm2

Can also be used for synthezising Si nanocrystal and SIMOX wafers! Implanter lay-out

MITopencourseware;M Schmidt,R O'Handley,S Ruff As a first order approximation for implantation; the impurity concentration as a  2 2 N(x)  e (x  R p ) / 2R p  function of depth in an amorphous 2 R p solid will be given by Gaussian distribution

Cp Rp is the projected Range and

Rp is the straggle

To calculate Rp one must determine the energy loss of the incident ion per unit length due to nuclear and electronic stopping interactions (LSS model)

S(E)=SN(E)+Se(E) Vertical projected range, Rp

The total distance the ion travels in the lattice is the range; R

The projected range, Rp is the average depth from the surface

http://www.cleanroom.byu.edu/rangestraggle.phtml Fd  v  E

MITopencourseware;M Schmidt,R O'Handley,S Ruff

Control of implant dose Q (number of /area); Ion current (and time) is measured by using a Faraday cup (guarded against secondary electron contribution). Very accurate ( error < 2% ) !

t I d Q  dt (At/cm2) d  0 nqA

Total implant energy; energy   power dt   IVdt  V  Idt  VQ Eq. 5.6 Boron profile: Pearson type IV distribution Boron is spreading more backwards, while heavy (eg As) spreading more forwards

Dose can be calculated as area under profile (Gaussian)= Q  2 C p R p Channeling and lateral projected range

If the ion velocity is parallel to a major orientation, some ion may travel considerable distance with little energy loss due to; a)nuclear stopping is not efficient (glancing elastic collisions) b)electron density is low Channeling can produce a significant tail in the profile! Channeling is characterized by the critical angle; Ψ (less angle higher risk for channeling) Z Z   9.73 i t Eq. 5.27 E0d IMPLANTATION DAMAGE

Lattice damage (and dopants activated) must be recovered via post-anneal SUMMARY:ImplantationImpurity solubility

The design of implantation systems

A model for calculation of implant parameters such as Rp,DRp Physical limitations (channeling, damages, shallow profiles) Simulations by; SRIM/TRIM/Synopsys etc

Strength: Precise control of dopant concentration Good dopant uniformity Good control of dopant penetration depth Produces a pure beam of ions Low temperature processing Ability to implant ions through films No solubility limit

Weakness: Very shallow or very deep profiles are difficult or impossible Incident ions damage must be repaired Throughput is limited Ion implant equipment is complex and hence very expensive, but also a great safety risk