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

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 Q BD . 2) Increase in tunneling current will J TAT+F.N > J F.N achieve a certain Q BD in a shorter T BD + T TAT+F.N < T F.N 3) Trap-Assisted-Tunneling (TAT) increases tunneling current.

Defects and breakdown in SiC • What defects will increase the tunneling current and result in a shorter lifetime? ± kT • A defect band can also result in TAT at a certain depth. • SiC : large density of defects above the band edge E D ± kT • The distribution of defects can be broad.

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 ± kT • The average distance between defects is > 10nm • The possibility of lined up defects is still low • It is possible not to consider multiple- TAT T ox

Defects and breakdown in SiC • How does one defect enhance the tunneling current locally? • TAT tunneling is a two-step process : the slower step determines the total probability ? ? T ox

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. x 0 • The sweet spot location is field dependent. ? T ox x 1

Defects and breakdown in SiC • At the sweet spot, the current enhancement coefficient is field dependent. x 0 ? T ox

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 ? T ox

Defects and breakdown in SiC • For Eox = 8.2 MV/cm: • For a sample of intrinsic SiO2/SiC capacitors T ox

Defects and breakdown in SiC • For Eox = 8.2 MV/cm: • What is the effect of one defect at the sweet spot on the weibull distribution • T bd = T bd 0 / η T ox

Defects and breakdown in SiC • For Eox = 8.2 MV/cm: • 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 T ox

Defects and breakdown in SiC • For Eox = 8.2 MV/cm: • What happens if we have a large number of defects at sweet spot : 1e6, 1e7, 1e8 ? • Area scaling statistics T ox

Defects and breakdown in SiC • For Eox = 8.2 MV/cm: • For a sample of intrinsic SiO2/SiC capacitors • What if the defects are not in the sweep spot ? T ox

Defects and breakdown in SiC • For Eox = 8.2 MV/cm: • What is the effect of one defect at 0.5nm from the sweet spot. • T bd = T bd 0 / η T ox

Defects and breakdown in SiC • For Eox = 8.2 MV/cm: • Area scaling for the defect 0.5nm away from sweet spot T ox

Defects and breakdown in SiC • For Eox = 8.2 MV/cm: • What happens if we have a large number of defects at 0.5 nm from sweet spot : 1e6, 1e7, 1e8 ? T ox

Defects and breakdown in SiC Eox = 8.2 MV/cm: • Effect of defects at sweet spot. • Effect of defects 0.5nm from sweet spot.

Defects and breakdown in SiC Eox = 8.2 MV/cm: • A distribution of defects in and away from the sweet spot will result in a continuum of failure distributions shorter than intrinsic. T ox

Defects and breakdown in SiC Eox = 6.2 MV/cm • At smaller fields, the lieftime continuum is more spread. Eox = 8.2 MV/cm • This effect has a limit beyond which lower failures are caused by traditional causes.

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.

Time Dependent Dielectric Breakdown E. Wu, IEEE Trans. Electron Devices 2002 • A uniformly reliable dielectric result in a Weibull distribution of failures