1. NBTI and Spin Dependent Charge Pumping in 4H-SiC MOSFETs
Mark A. Anders, Patrick M. Lenahan,
Pennsylvania State University
Aivars Lelis, US Army Research Laboratory

2. Physical Basis for ESR (Simple Case) At resonance, spin flips
Deviations from the resonance condition provide useful information about the nature of specific defects
Energy
Magnetic Field
Electron-Nuclear Hyperfine Interaction: due to nearby nucleus with magnetic moment
g bH + mIA
Energy
Magnetic Field
Spin-Orbit Coupling: due to electron’s orbital angular momentum about the nucleus
g bH
At resonance, spin flips

3. Spin-Orbit Coupling
In the data shown, the factor most responsible for deviations from hv=geβH
Bohr model: electron orbits nucleus;
angular momentum n h (units of h).
To an observer on the electron, it looks
like the nucleus is orbiting the electron.
The greater the nuclear charge and
orbital angular momentum quantum
number, the greater the spin orbit
coupling. The orbiting nucleus looks like
a current loop generating a magnetic
field. This contribution is expressed in the g tensor.

4. Electrically Detected Magnetic Resonance (EDMR)
Problem
Conventional EPR has a sensitivity of about 1010 total paramagnetic defects It is also sensitive to ALL paramagnetic defects in a sample
We want to identify defects in transistors
We want to know what different defects do to device performance
A main problem for electronic materials science is performing resonance inside fully processed transistors in integrated circuits
EDMR provides sensitivity about 7 orders of magnitude higher than conventional EPR
Solution: EDMR, spin dependent recombination (SDR),
and spin dependent trap-assisted tunneling (SDT)

5. Spin Dependent Recombination (SDR): Shockley-Read-Hall Model

6. Pauli Exclusion Principle
EDMR via SDR:
Pauli Exclusion Principle

7. Background: What are these defects?
Anticipated spectrum of the negatively charged silicon vacancy with the field parallel to the c-axis utilizing the hyperfine parameters of Isoya et al.
J. Isoya, T. Umeda, N. Mizuochi, N. T. Son, E. Janzen, and T. Ohshima, Phys. Status Solidi B 245(7), 1298 (2008). 7

8. Background: Silicon Vacancies
Silicon Vacancy Model
Experimental Results

9. Previous Work: NBTI x5 Hydrogen or Nitrogen related defect
Pre-stress
190oC
Post-stress
Hydrogen or Nitrogen related defect
At interface, bulk, or possibly oxide
Possible nitrogen related defect
At interface or bulk

10. NBTI: “Bulk Defects” +++++
Utilizing VDMOSFET geometry, we probe bulk defects
Biasing scheme:
Strongly accumulate channel
Forward bias drain/body diode
“Bulk” recombination dominant current N+ P
Source i
-Vg
Gate P+ Vd
Drain
N+ SiC substrate
N SiC Epi
Source
Gate
+++++ N+
accumulated channel P+ P

11. Low Frequency EDMR
150, or 250, or 360 MHz (low field) vs. ~9.5 GHz (X-band)
Broadening due to g is field/frequency dependent
Broadening due to hyperfine interactions is field/frequency independent (to first order)
Simulated using easyspin

12. NBTI: Hydrogen Movement in Bulk
X-band
Low field
Previously, X-band measurements suggests broadening due to g or hyperfine
EDMR measurements at low field reveal that broadening is at least partially due to hyperfine interactions

13. The Fundamental Problem
EDMR via SDR:
The Fundamental Problem
This technique is appropriate for levels around mid-gap.
It does not, however, address traps near the band edges.
We are interested in what’s happening near the conduction band and valence band.
Spin dependent charge pumping: a solution.

14. Spin Dependent Charge Pumping (SDCP)
By utilizing a pulse wave at the gate voltage, we can explore more of the band gap with larger amplitude.

15. 3-level Energy Resolved Charge Pumping
semiconductor
oxide G D S
Waveform Generator 𝑖 … a b c
time
Gate Voltage
(Vary b Voltage)
Flood interface with holes to recombine with interface trap electrons
(a)
(b)
(c) EF EF EF
Fill Traps
Allow time for traps above EF to empty

16. Enhanced sensitivity (2000x greater)
Recap: SDCP
Enhanced sensitivity (2000x greater)

17. SDCP (full bandgap) vs BAE (midgap)
NO-annealed pMOSFET
Probing states throughout most of the bandgap shows a different line shape and broader center line compared to only mid-gap states

18. SDCP (full bandgap) vs BAE (midgap)
The “full bandgap” spectra have g-values much lower than the “mid-gap” spectra in this sample
Mid gap (BAE) spectra are mostly due to silicon vacancy

19. NO vs non-NO 350 MHz ~9.5 GHz C-axis is parallel to magnetic field
c-axis parallel to magnetic field
c-axis parallel to magnetic field
C-axis is parallel to magnetic field
NO annealed nMOSFET has broader shoulders than an nMOSFET which did not receive an NO anneal
Many abundant magnetic nuclei, likely nitrogen, would cause broadening

20. NO vs non-NO
350 MHz
~9.5 GHz
NO annealed nMOSFET has a broader center line at both high and low fields
Many abundant magnetic nuclei, likely nitrogen, would cause broadening

21. NO vs non-NO
~9.5 GHz
The NO annealed device spectra is much broader than the non-NO device spectra
More evidence for Nitrogen hyperfine interactions

22. Carbon Dangling Bonds? C dangling bond: g//c = 2.0023 g⊥c = 2.0032
J. L. Cantin et al.
If carbon dangling bonds largely contributed to the interface state density, we would observe a line at g//c = and g⊥c = We do not. There are no carbon dangling bonds at the interface.

23. 3-level SDCP
Non-NO nMOSFET

24. 3-level SDCP Non-NO nMOSFET Side peak amplitudes are reduced
Defect responsible for side peaks are not near the conduction band edge

25. Frequency-dependent SDCP
w/out NO nMOSFET
NO nMOSFET
Lower charge pumping frequencies probe further into the oxide
Could provide a way to look a little into the gate oxide
More work needs to be done to fully understand multi-frequency SDCP results

26. Prelim Results on GaN Low -field EDMR: GaN p-n diode
The complex spectra is clearly due to hyperfine interactions

27. Prelim Results on GaN EDMR Amplitude vs Gate Bias
EDMR amplitude of several different spectrum peaks
Calculated recombination current