EE 515

Diffusion 4

Arun Tej M.
D = f(N)
Lateral Diffusion
Doping Concentration Dependence
• Fick’s first law is based on the premise that flow is directly proportional to the concentration
gradient
• Actual concentration profile of many diffused dopants looks box-like i.e., diffusion appears faster at
higher concentrations

Boron diffusion profiles (erfc) for constant

surface concentration
• Higher concentration constant surface concentration profile → not well modeled by a simple erfc
profile with constant 𝐷
• Reason: Diffusivity varies with concentration with higher-concentration regions diffusing faster and
smoothing the concentration profile
• Fick’s first law: Only gradient dependence – Not on absolute value
 Needs modification

• Fick’s formula with non-constant 𝐷 – Usually cannot be integrated

directly – Numerical solutions
𝜕𝐶 𝜕 𝜕𝐶
= 𝐷
𝜕𝑡 𝜕𝑥 𝑒𝑓𝑓 𝜕𝑥

• Experimental data is found to be well modeled using a diffusivity that is proportional to carrier
concentration 𝐷 ∝ 𝑛 𝑛𝑖 or 𝐷 ∝ 𝑛 𝑛𝑖 2

𝑛 − 𝑛2 2− 𝑝 + 𝑝2 2+
𝐷𝑒𝑓𝑓 = 𝐷𝑖0 + 𝐷𝑖 + 2 𝐷𝑖 (for n-type dopants) 𝐷𝑒𝑓𝑓 = 𝐷𝑖0 + 𝐷𝑖 + 2 𝐷𝑖 (for p-type dopants)
𝑛𝑖 𝑛𝑖 𝑛𝑖 𝑛𝑖
Charged Vacancies
• Vacancies in Si exist in the charge states 𝑉 + , 𝑉 − and 𝑉 2− (in addition to neutral 𝑉 0 )

Si 𝐸𝐶
𝑉 0 + ℎ+ ⇌ 𝑉 + Hole trap 0.11 𝐸 2−
0.44
𝑉0 + 𝑒− ⇌ 𝑉− 𝐸−
Electron trap
𝑉0 + 2𝑒 − ⇌ 𝑉 2−
𝐸+
0.06 − 0.12
𝐸𝑉

• The impurities (𝐼) get attached with charged vacancy states and form pairs such as
Boron: 𝐼 − 𝑉 + and 𝐼 − 𝑉 0
Phosphorous: 𝐼 − 𝑉 2− , 𝐼 − 𝑉 − , and 𝐼 − 𝑉 0
Arsenic: 𝐼 − 𝑉 2− , 𝐼 − 𝑉 − , and 𝐼 − 𝑉 0 (Advantage: low diffusivity - shallow junctions, high solid solubility)
• These pairs can diffuse and may contribute to additional diffusivity
Charged Vacancies
• Total vacancy concentration 𝑉 = 𝑉 0 + 𝑉 + + 𝑉 − + 𝑉 2−
• 𝑉 0 → Unaffected by any electric field due to impurity concentration gradients ⇒ 𝑉 0 will remain
same through out the material ⇒ 𝑉 0 in extrinsic material is equal to that in intrinsic material
• 𝑉 0 - a function of temperature alone
• Concentrations of various charge states of the impurity are related by the position of Fermi level
• 𝑉 - a function of the impurity concentration
• Charge interactions in a solid can be treated in the same way as chemical interactions in solutions
• Equilibrium constant for 𝑉 + is written as: 𝑉+
𝑘1 =
𝑉 0 [ℎ]
𝑉𝑖+
• If, at the diffusion temperature, Si remains intrinsic, then 𝑘1 =
𝑉𝑖0 𝑛𝑖
𝑉𝑒+
• If, at the diffusion temperature, Si becomes extrinsic, then 𝑘1 =
𝑉𝑒0 𝑝
 𝑉𝑖+ 𝑉𝑒+ 𝑉𝑖+ 𝑝
𝑘1 = = ⇒ 𝑉𝑒+ = 𝑉𝑖 = 𝑉𝑖0 + 𝑉𝑖+ + 𝑉𝑖− + 𝑉𝑖2− ⇒
0
𝑉𝑖 𝑛𝑖 𝑉𝑒0 𝑝 𝑛𝑖

𝑉𝑖− 𝑛 𝑉𝑖2− 𝑛2 𝑝 + 𝑛 − 𝑛2 2−
Similarly, 𝑉𝑒− = 𝑉𝑒2− 𝑉𝑒 = 𝑉𝑖0 + 𝑉 + 𝑉 + 2 𝑉𝑖
= 𝑛𝑖 𝑖 𝑛𝑖 𝑖 𝑛𝑖
𝑛𝑖 𝑛𝑖2

If the interactions of impurities with vacancies of various charged states are independent of each other,
in an analogous form, diffusivity can be written as

𝑝 + 𝑛 − 𝑛2 2−
𝐷𝑖 𝑒𝑓𝑓 = 𝐷𝑖0 + 𝐷𝑖+ + 𝐷𝑖− + 𝐷𝑖2− 𝐷𝑒 𝑒𝑓𝑓 = 𝐷𝑖0 + 𝐷 + 𝐷 + 2 𝐷𝑖
𝑛𝑖 𝑖 𝑛𝑖 𝑖 𝑛𝑖
2
𝑛 𝑛
𝐷𝑒𝑓𝑓 = 𝐷𝑖0 + 𝐷𝑖− + 2 𝐷𝑖2− for n-type dopants
𝑝 + 𝑛 − 𝑛2 𝑛𝑖 𝑛𝑖
𝐷𝑒𝑓𝑓 = 𝐷𝑖0 + 𝐷𝑖 + 𝐷𝑖 + 2 𝐷𝑖2−
𝑛𝑖 𝑛𝑖 𝑛𝑖 𝑝 + 𝑝2 2+
𝐷𝑒𝑓𝑓 = 𝐷𝑖0 + 𝐷𝑖 + 2 𝐷𝑖 for p-type dopants
𝑛𝑖 𝑛𝑖
 Concentration dependence
 𝑛 𝑛 2
𝐷𝑒𝑓𝑓 = 𝐷𝑖0 + 𝐷𝑖− + 2 𝐷𝑖2− 𝐷𝑖 = 𝐷𝑖0 + 𝐷𝑖− + 𝐷𝑖2−
𝑛𝑖 𝑛𝑖

Each of these individual diffusivities can be written in an Arrhenius form 𝐷 = 𝐷𝑜 𝑒 −𝐸𝑑/𝑘𝑇

𝑛 𝑛2
1+𝛽 +𝛾 2
𝑛𝑖 𝑛𝑖 𝐷𝑖− 𝐷𝑖2−
𝐷𝑒𝑓𝑓 = 𝐷𝑖 where 𝛽= 0 and 𝛾= 0
1+𝛽+𝛾 𝐷𝑖 𝐷𝑖

𝛽- factor represents the linear variation in the diffusion coeff. in highly doped material
𝛾- factor represents the quadratic variation in the diffusion coeff. in highly doped material
𝐷𝑜
𝐸𝑑
 Example
Calculate the effective diffusion coefficient at 1000 𝑜𝐶 for two different box-shaped As profiles grown
by Si epitaxy, one doped at 1 × 1018 𝑐𝑚−3 and the other doped at 1 × 1020 𝑐𝑚−3 . At 1000 𝑜𝐶 , the
intrinsic concentration is 7.14 × 1018 𝑐𝑚−3 .

𝐷𝑜 = 0.011 𝑐𝑚2 𝑠 −1 𝐸𝑑 = 3.44 𝑒𝑉

𝐷𝑜− = 31.0 𝑐𝑚2 𝑠 −1 𝐸𝑑− = 4.15 𝑒𝑉

𝐷𝑖 = 𝐷𝑖0 + 𝐷𝑖− + 𝐷𝑖2− 𝐷 = 𝐷𝑜 𝑒 −𝐸𝑑/𝑘𝑇

3.44 4.15
− −
= 0.011𝑒 𝑘×1273 + 31.0𝑒 𝑘×1273

= 2.67 × 10−16 + 1.17 × 10−15

= 1.43 × 10−15 𝑐𝑚2 𝑠 −1
Intrinsic Diffusion Coefficients of Dopants in Si
• 𝑛𝑖 at 1000 𝑜𝐶 is 7.14 × 1018 𝑐𝑚−3 𝐷 0 (𝑐𝑚2 𝑠 −1 ) 𝐸𝐴 (𝑒𝑉)
• Intrinsic for doping levels less than 𝑛𝑖 𝑆𝑖 560 4.76
𝐵 1.00 3.50
𝐼𝑛 1.20 3.50 Fast diffusers
𝑃 4.70 3.68
𝐴𝑠 9.17 3.99
Slow diffusers
𝑆𝑏 4.58 3.88

Different values for extrinsic conditions

• Shallower junctions in recent times
→ Slow diffusers have become more important
→ 𝐴𝑠 has become choice for n-type doping
• 𝐵 for p-type doping because of its high solid solubility
𝐷 = 𝐷 0 𝑒 −𝐸𝐴 𝑘𝑇 • Shallow junctions with p-type is relatively difficult
Fitting a single activation energy under intrinsic conditions, we get

𝐷 0 (𝑐𝑚2 𝑠 −1 ) 𝐸𝐴 (𝑒𝑉)
−3.99/𝑘×1273
𝐷𝑖 = 9.17𝑒 𝑆𝑖 560 4.76
𝐵 1.00 3.50
= 1.48 × 10−15 𝑐𝑚2 𝑠 −1
𝐼𝑛 1.20 3.50
𝑃 4.70 3.68
𝐴𝑠 9.17 3.99
𝑆𝑏 4.58 3.88
Example
Calculate the effective diffusion coefficient at 1000 𝑜𝐶 for two different box-shaped As profiles grown
by Si epitaxy, one doped at 1 × 1018 𝑐𝑚−3 and the other doped at 1 × 1020 𝑐𝑚−3 . At 1000 𝑜𝐶 , the
intrinsic concentration is 7.14 × 1018 𝑐𝑚−3 .

𝐷𝑜 = 0.011 𝑐𝑚2 𝑠 −1 𝐸𝑑 = 3.44 𝑒𝑉

𝐷𝑜− = 31.0 𝑐𝑚2 𝑠 −1 𝐸𝑑− = 4.15 𝑒𝑉

𝑛 𝐷 = 𝐷𝑜 𝑒 −𝐸𝑑/𝑘𝑇
𝐷𝑖 = 𝐷𝑖0 + 𝐷𝑖− + 𝐷𝑖2−
𝑛𝑖
3.44
−𝑘×1273
4.15
−𝑘×1273 1 × 1020
= 0.011𝑒 + 31.0𝑒
7.14 × 1018
= 2.67 × 10−16 + 1.63 × 10−14

= 1.66 × 10−14 𝑐𝑚2 𝑠 −1 Ten-fold higher for higher doping case

 • Away from the surface, doping concentration (say
phosphorous) decreases
𝑛+
Expected • Close to the E-B junction, the concentration goes
𝑝 below intrinsic concentration ⇒ 𝑛 < 𝑛𝑖
base width
• As the diffusion progresses, the impurity-vacancy
𝑛 pairs, which form at high concentrations, will be
released.

• These vacancies move down towards C-B junction

𝐼. 𝑉 − → 𝐼 + 𝑉 −
→ 𝐼− + 𝑉 0 • Boron atoms now have additional places to go into

• Higher junction depth i.e., higher base width

Special Problems in Silicon Diffusion
• Redistribution during oxide growth
• Diffusion during oxide growth
• Cooperative diffusion effects
Redistribution During Oxide Growth
• During IC fabrication, diffusions are made through windows cut in a mask on the surface of the
silicon
• Most convenient mask – SiO2, grown out of Si itself by thermal oxidation
• This can be done as a separate step after a diffusion
• Or, it can be done during a two-step diffusion process where oxide is simultaneously grown during
the drive-in phase (→ preferable method since it saves additional masking step)
• In either case, some of the Si (that is already impurity-doped) is consumed to form the oxide
• This results in a redistribution of impurities
• Recall lectures on thermal oxidation: Advanced Point Defect Based Models
Redistribution During Oxide Growth
• During IC fabrication, diffusions are made through windows cut in a mask on the surface of the
silicon
• Most convenient mask – SiO2, grown out of Si itself by thermal oxidation
• This can be done as a separate step after a diffusion
• Or, it can be done during a two-step diffusion process where oxide is simultaneously grown during
the drive-in phase (→ preferable method since it saves additional masking step)
• In either case, some of the Si (that is already impurity-doped) is consumed to form the oxide
• This results in a redistribution of impurities
• Recall lectures on thermal oxidation: Advanced Point Defect Based Models
• The extent of redistribution is a function of
• Rate at which Si is consumed
• Relative diffusivities and solid solubilities of the impurities in Si and SiO2
Redistribution During Oxide Growth
• Distribution Coefficient 𝑚- Ratio of the equilibrium concentrations of impurity in Si and in SiO2
• 𝑚 ~ 20 for Ga; 𝑚 ~ 10 for P, As, and Sb
• 𝑚 for B is crystal orientation dependent
• 𝑚 ≈ 33 𝑒 −0.52/𝑘𝑇 for (100) orientation (typically between 0.15 to 0.3)
• 𝑚 ≈ 20 𝑒 −0.52/𝑘𝑇 for (111) orientation
• For 𝑚 < 1, the growing oxide takes up the impurity from Si
• ⇒ For B, impurity depletion occurs at the interface on the side of Si
• Extent of this depletion also depends on the rate at which B diffuses through SiO2 and
escapes
• Diffusion rate of B in SiO2 is orders of magnitude lower than that in Si.
• Redistribution effects dominate
Redistribution During Oxide Growth
• Distribution Coefficient 𝑚- Ratio of the equilibrium concentrations of impurity in Si and in SiO2
• 𝑚 ~ 20 for Ga; 𝑚 ~ 10 for P, As, and Sb
• 𝑚 for B is crystal orientation dependent
• 𝑚 ≈ 33 𝑒 −0.52/𝑘𝑇 for (100) orientation (typically between 0.15 to 0.3)
• 𝑚 ≈ 20 𝑒 −0.52/𝑘𝑇 for (111) orientation
• For 𝑚 = 1, some depletion still occurs because
• Si roughly doubles in volume on forming SiO2
• Thus, some diffusion will occur from the highly doped Si into the relatively lightly doped oxide
Redistribution During Oxide Growth
• Distribution Coefficient 𝑚- Ratio of the equilibrium concentrations of impurity in Si and in SiO2
• 𝑚 ~ 20 for Ga; 𝑚 ~ 10 for P, As, and Sb
• 𝑚 for B is crystal orientation dependent
• 𝑚 ≈ 33 𝑒 −0.52/𝑘𝑇 for (100) orientation (typically between 0.15 to 0.3)
• 𝑚 ≈ 20 𝑒 −0.52/𝑘𝑇 for (111) orientation
• For 𝑚 > 1, the growing oxide rejects the impurity in Si
• Impurity pileup is expected to occur at the interface on the side of Si
• Extent of this pileup also depends on the rate at which it diffuses through SiO2 and escapes
• Diffusion rate of P in SiO2 is slow. Thus, pileup occurs.
• Diffusion rate of Ga in SiO2 is very fast. End result is still a depletion.
• In practice, pileup effect of P is relatively small and can be neglected – Redistribution and
Diffusion effects cancel out in this case.