Defects Impact on Time dependent dielectric breakdown in SiC MOSFET - - PowerPoint PPT Presentation

Defects Impact on Time dependent dielectric breakdown in SiC MOSFET

• Z. Chbili , K. P. Cheung

Semiconductor Electronics Division, NIST, Gaithersburg, MD

Motivation

• S. Arthur, ARL workshop 2013
• 1. High density of defects in SiO2/SiC.
• 2. TDDB: Early and “extrinsic” failures

are a serious reliability issue in SiC

Motivation

• Collective TDDB data:
• ~ 500 devices
• All stress conditions

normalized

• What are these early failures, and what are the available solutions?

Motivation

• Traditional causes of early failures:
• Particles, Contamination, Local

thinning of the oxide …

• Traditional solutions:
• 1. Clean-up the process
• 2. Screening

Motivation

• Screening in Weibull distribution:
• Possibility of screening
• Impossible screening

` `

Motivation

• What if the cause of early failures in SiC is different?
• Do the traditional solutions still work ?

TDDB background

well known TDDB facts: 1) For thick oxides, there is a critical charge to breakdown QBD. 2) Increase in tunneling current will achieve a certain QBD in a shorter TBD 3) Trap-Assisted-Tunneling (TAT) increases tunneling current.

JTAT+F.N > JF.N

TTAT+F.N< TF.N

+

Defects and breakdown in SiC

• What

defects will increase the tunneling current and result in a shorter lifetime?

• A defect band can also result in TAT at a

certain depth.

• SiC : large density of defects above the

band edge

• The distribution of defects can be

broad. ± kT ED ± kT

Defects and breakdown in SiC

• For a uniform distribution of defects

1e13 cm-2 eV-1

• We only consider ~ 5e11 cm2 defects

above the band edge

• The average distance between defects

is > 10nm

• The possibility of lined up defects is still

low

• It is possible not to consider multiple-

TAT Tox ± kT

Defects and breakdown in SiC

• How does one defect enhance the

tunneling current locally? Tox

? ?

• TAT tunneling is a two-step process :

the slower step determines the total probability

Defects and breakdown in SiC

• There is a sweet spot where the

probability of tunneling through each barrier are equal

• The additional tunneling current is

maximum if the defect is at the sweet spot.

• The sweet spot location is field

dependent. Tox

?

x0 x1

Defects and breakdown in SiC

• At

the sweet spot, the current enhancement coefficient is field dependent. Tox

?

x0

Defects and breakdown in SiC

• For Eox = 8.2 MV/cm:
• X (η) = 1.025nm
• η = 2.5e4
• Need to consider the avg

around the size of the defect wavefunction:

• η! = 8e3

Tox

?

Defects and breakdown in SiC

• For

a sample

• f

intrinsic SiO2/SiC capacitors Tox

• For Eox = 8.2 MV/cm:

Defects and breakdown in SiC

• What is the effect of one defect

at the sweet spot on the weibull distribution

• Tbd = Tbd0/ η

Tox

• For Eox = 8.2 MV/cm:

Defects and breakdown in SiC

• But the 1 defect effect is only

local : we should consider area scaling around the area of defect.

• weakest link.
• We consider the size of the

defect to be roughly 1nm2 Tox

• For Eox = 8.2 MV/cm:

Defects and breakdown in SiC

• What happens if we have a large

number of defects at sweet spot : 1e6, 1e7, 1e8 ?

• Area scaling statistics

Tox

• For Eox = 8.2 MV/cm:

Defects and breakdown in SiC

• For

a sample

• f

intrinsic SiO2/SiC capacitors

• What if the defects are not in

the sweep spot ? Tox

• For Eox = 8.2 MV/cm:

Defects and breakdown in SiC

• What is the effect of one defect

at 0.5nm from the sweet spot.

• Tbd = Tbd0/ η

Tox

• For Eox = 8.2 MV/cm:

Defects and breakdown in SiC

• Area scaling for the defect

0.5nm away from sweet spot Tox

• For Eox = 8.2 MV/cm:

Defects and breakdown in SiC

• What happens if we have a large

number of defects at 0.5 nm from sweet spot : 1e6, 1e7, 1e8 ? Tox

• For Eox = 8.2 MV/cm:

Defects and breakdown in SiC

• Effect of defects at sweet spot.

Eox = 8.2 MV/cm:

• Effect of defects 0.5nm from

sweet spot.

Defects and breakdown in SiC

• A distribution of defects in and away from

the sweet spot will result in a continuum of failure distributions shorter than intrinsic. Tox Eox = 8.2 MV/cm:

Defects and breakdown in SiC

• At smaller fields, the lieftime

continuum is more spread.

• This effect has a limit beyond

which lower failures are caused by traditional causes.

Eox = 6.2 MV/cm Eox = 8.2 MV/cm

Motivation

• Can trap assisted tunneling explain the broad failure distribution in our

collective TDDB data?

• If this is the case, screening is impossible.

Early failure screening

• Can trap assisted tunneling explain the broad failure distribution in our

collective TDDB data?

• If this is the case, screening is impossible.

Early failure screening

• Can trap assisted tunneling explain the broad failure distribution in our

collective TDDB data?

• If this is the case, screening is impossible.

Early failure screening

• Can trap assisted tunneling explain the broad failure distribution in our

collective TDDB data?

• If this is the case, screening is impossible.

Conclusion

• We presented enough indications that TAT in SiO2 defects could be the

cause of early failures in SiC MOS devices.

• Such early failures are not possible to screen out.
• This problem will not be possible to solve by traditional means of

“cleaning” the oxidation process.

• A new method of oxide growth should be adopted.

• E. Wu, IEEE Trans. Electron Devices 2002

Time Dependent Dielectric Breakdown

• A uniformly reliable dielectric result in a Weibull distribution of failures