Direct Experimental Evidence Linking Silicon Dangling Bond Defects to Oxide Leakage Currents

1
Direct Experimental Evidence Linking Silicon
Dangling Bond Defects to Oxide Leakage Currents
P.M. Lenahan, J.J. Mele, A.Y. Kang, J. P.
Campbell Penn State University, University Park,
PA 16802   S.T. Liu Honeywell Corp. Plymouth, MN
55441   R.K. Lowry and D. Woodbury Intersil
Corp. Melbourne, FL 32902 R. Weimer, Micron
Technologies Boise, Idaho 83707-0006
2
Ec
Stress Induced Leakage (SILC)
Trap
CB Edge
Inelastic tunneling of silicon conduction band
electrons through oxide defects near the Si/SiO2
boundary.
EF
VB Edge
CB Edge
Ev
SiO2
Si
VB Edge
S.Takagi, et al. Trans.Electron.Dev 46, 348 (1999)
E.Rosenbaum and L.F.Register, IEEE
Trans.Electron.Dev. 44, 317(1997)
3
Literature suggests oxygen vacancy centers (E
centers)

• J.H.Suehle, et al. IRPS (1994)
• J.H.McPherson and H.Mogul, J.Appl.Phys, 84, 1513
  (1998)
• B.Schlund, et al. IRPS (1996)
• D.J.Dumin and J.Maddux, IEEE Trans.Electron.Dev.,
  40, 886 (1993)
• S.Takagi, et al. IEEE Trans.Electron.Dev., 46,
  348 (1999)
• A.Yokozawa, et al. IEDM (1997)

4
(No Transcript)
5
At least two independent studies indicate that E
centers are generated when oxides are subjected
to high electric fields.   So, if we could
link E center density to leakage current, we
could establish an important role for the centers
in oxide leakage phenomena. W. L. Warren and
P.M. Lenahan, JAP 62, 4305 (1987), IEEE Trans
Nucl. Sci 34, 1355 (1987) H. Hazama, et al.
Proceedings of the Workshop on Ultra Thin Oxides
Jap.Soc. Appl. Phys. Tokyo 1998. Cat. No. Ap
982204 pp201-212.
6
Test E hypothesis with neutral E centers
Ec
Ec
Ec
Ec
Ev
Ev
Ev
Ev


SiO2
Si
Si
SiO2
(Any net space charge near the Si/SiO2 boundary
will decrease the tunneling barrier and increase
oxide current for any gate potential.)
7
Approach (a) Generate neutral E centers in a
wide variety of oxides, annihilate the E
centers by various means.
(b) Compare generation and annihilation of the E
centers with generation and annihilation of oxide
leakage current. (Are they strongly correlated?)
(c) Compare experimental results and theoretical
work on inelastic tunneling and SILC. (Are the
defect densities reasonable in terms of the
(very crude) theory available.)
8
Oxides Utilized in the Study   3.3nm (forming
gas) 3.3nm (no forming gas) 45 nm (forming
gas)   (all thermally grown)
9
(No Transcript)
10
3.3nm Oxide (forming gas)
Arbitrary Units
As Processed
Post VUV
Post Anneal
Pb0
E
3440
3445
3450
3455
3460
3465
3470
3475
3480
3485
Magnetic Field (Gauss)
11
(No Transcript)
12
E and Leakage Current Generation 3.3nm oxide
(forming gas)
13
E and Leakage Current Anneal 3.3nm oxide
(forming gas)
14
3.3nm Oxide (no forming gas)
90 MIN ANNEAL
90 MIN VUV
Virgin
ESR Amplitude (Arb. Units)
Pb0
E
3440
3450
3460
3470
3480
Magnetic Field (G)
14
15
I-V Characteristics of 3.3nm Oxide (no forming
gas)
16
E and Leakage Current Generation 3.3nm oxide (no
forming gas)
17
E and Leakage Current Anneal 3.3nm oxide (no
forming gas)
18
45nm Oxides
As Processed
Post VUV
Arbitrary Units
Post VUV and Anneal
Pbo (g2.0058)
E g(z.c.)2.0005
3440.5
3445.5
3450.5
3455.5
3460.5
3465.5
3470.5
3475.5
3480.5
Magnetic Field (G)
19
(No Transcript)
20
E and Leakage Current Generation 3.3nm oxide
(forming gas)
20
21
E and Leakage Current Anneal 45nm oxide (forming
gas)
22
Conclusions
In several quite different oxides we find that
(1) Generation of E centers is accompanied by
generation of oxide leakage currents.
(2) A brief 200oC anneal in air annihilates most
of the E centers and most of the leakage current.
23
Since (A) earlier work by two independent groups
show that E centers are generated by high field
stressing oxides, And (B) recent theoretical
and experimental studies indicate that E centers
are good candidates for leakage current defects,
we conclude that E centers are important,
probably dominating defects in SILC (and RILC) in
a wide range of oxides on silicon.
24
Many studies report generation of interface
states in conjunction with the creation of stress
induced leakage currents. Several studies also
report a strong correlation between SILC and
interface state generation. Why is this so?
25
Before Stressing
Oxide
Si
Si
Si
Si
Si
Si
Si
Si
H
H
H
H
Si/SiO2
Si
Si
Si
Si
After Stressing
Oxide
Si
Si
Si
Si
H
H
H
H
Si/SiO2
Si
Si
Si
Si
26
Consider Statistical MechanicsThe system will
approach the lowest Gibbs Free Energy
G H-TS
E
EH
O
O
Si
O
Si
O
H
Oxide
O
O
Pb
PbH
H
Si
Si
Si/SiO2
Si
Si
Si
Si
Si
Si
h
0
(The oxide and interface Si-H bond enthalpies
will be about equal)
27
Entropy S k ln (O)
(Total of N sites)
Suppose all E dbs are unpassivated

O
O
O
Si
O
Si
Si
O
O
O
O
O
S k ln(1)
Contribution to configurational entropy of N E
sites
Suppose all Pb dbs are passivated
(Total of M sites)

H
H
H
H
Si
Si
Si
Si
Contribution to configurational entropy of M PbH
sites S k ln(1)
28
Suppose we remove one H from the M PbH sites
the configurational entropy changes ?S k ln (M)

H
H
H
Si
Si
Si
Si
Suppose we add one H to the N E sites the
configurational entropy changes ?S k ln (N)

O
O
O
Si
Si
Si
H
O
O
O
O
O
O
The Gibbs free energy of the system will be
lowered by the transfer of some hydrogen from Pb
sites to E sites. (If kinetics allows it)
29
The process will occur to some extent.
To how great an extent?
PbH E H2
Pb EH H2
Pb EH
exp ( - ?G / kT) K 1
PbH E