Noninvasive nature of corona charging on thermal Si/SiO2 structures

Article abstract of DOI:10.1063/1.1789576

The application of corona charging technique in thermal Si/SiO₂ semiconductor devices and electron spin resonance (ESR) experiments was analyzed. Investigation shows that the corona charging techniques were inherently unreliable and invasive. I

Full text of DOI:10.1063/1.1789576

Noninvasive nature of corona charging on thermal Si ∕ Si O 2 structures
M. S. Dautrich, P. M. Lenahan, A. Y. Kang, and J. F. Conley Jr.
Citation: Applied Physics Letters 85, 1844 (2004); doi: 10.1063/1.1789576
View online: http://dx.doi.org/10.1063/1.1789576
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/85/10?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Paramagnetic point defects in ( 100 ) Si ∕ La Al O 3 structures: Nature and stability of the interface
J. Appl. Phys. 102, 034516 (2007); 10.1063/1.2749423
Invasive nature of corona charging on thermal Si/SiO 2 structures with nanometer-thick oxides revealed by
electron spin resonance
Appl. Phys. Lett. 82, 2835 (2003); 10.1063/1.1540245
Thermally induced interface degradation in (100) and (111) Si/SiO 2 analyzed by electron spin resonance
J. Vac. Sci. Technol. B 16, 3108 (1998); 10.1116/1.590449
Hydrogen-induced thermal interface degradation in (111) Si/SiO 2 revealed by electron-spin resonance
Appl. Phys. Lett. 72, 2271 (1998); 10.1063/1.121335
Electron spin resonance features of interface defects in thermal (100) Si/SiO 2
J. Appl. Phys. 83, 2449 (1998); 10.1063/1.367005
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
155.247.166.234 On: Sun, 23 Nov 2014 01:52:45

APPLIED PHYSICS LETTERS
VOLUME 85, NUMBER 10
6 SEPTEMBER 2004
Noninvasive nature of corona charging on thermal Si/SiO2 structures
M. S. Dautrich,a) P. M. Lenahan, and A. Y. Kang
Pennsylvania State University, University Park, Pennsylvania 16802
J. F. Conley, Jr.
Sharp Labs of America, Camas, Washington 98607
(Received 10 March 2004; accepted 7 July 2004)
The corona charging technique is widely utilized in commercial Si/SiO2 semiconductor device
reliability characterization tools and has been used in numerous electron spin resonance (ESR)
experiments, by several groups to study defect centers in Si/SiO2 system. A recent ESR study
argued that the corona charging approaches are inherently unreliable and invasive. In this work we
show that this is not the case. We find that low-field corona biasing is essentially noninvasive and
thus can be utilized in both reliability characterization and fundamental studies of defect
structures. © 2004 American Institute of Physics. [DOI: 10.1063/1.1789576]
Comparatively low-field corona biasing is an essential
part of several widely utilized commercial device reliability
characterization tools. Examples include the Quantox sys-
tems of KLA-Tencor1 and the Corona Oxide Characteriza-
tion of Semiconductor (COCOS) technology of Semiconduc-
tor Diagnostics.2 In several early studies involving electron
spin resonance (ESR),3–7 low-field corona biasing was uti-
lized to manipulate the position of the Fermi energy at the
Si/SiO2 boundary, providing information about the density
of states of Si/SiO2 interface trapping centers. In all of these
cases, the validity of the measurements requires that little or
no relevant damage occurs to the Si/SiO2 system as a result
of the biasing. Other early ESR studies,8,9 utilizing quite high
field corona bias, demonstrated that high oxide field corona
biasing can clearly generate both large densities of Si/SiO2
interface and oxide paramagnetic centers.
prior to oxidation. Room temperature magnetic resonance
measurements were made at X band with a commercial state-
of-the-art ESR spectrometer (Bruker Instruments) using a
TE104 double cavity and a weak pitch spin standard. Corona
biasing followed the scheme described by Weinberg et al.12
Low-field oxide bias was measured using a Kelvin probe. We
subjected devices to up to 20 h of corona biasing at various
oxide potentials, periodically making ESR measurements of
the oxide/silicon structures. The oxides were stressed with
the Si surface (n-type) in accumulation.
Figure 1 illustrates ESR traces taken on the thick
͑634 Å͒ oxides subjected to either high-field ͑+7 MV/cm͒
or lower-field ͑+4 MV/cm͒ corona biasing (a trace of an
unstressed sample has been included for reference). The
high-field stressed sample exhibited large numbers ͑Ϸ4.5
ϫ1011/cm2͒ of P and E centers. In the samples subjected
Ј
b
Recently, Stesmans and Afanas’ev10 have argued that co-
rona charging techniques are “inherently unreliable” and
that, when applied to thin oxides on silicon, the technique is
invasive and “intolerably modifies the inherent properties of
the entity studied, leaving it quite disrupted in terms of
(ESR-active) defects.” They argued that earlier ESR studies
done at low-field3–7,11 and the commercial device character-
ization tools are unreliable.
to equal periods of lower field corona biasing, we observed
near negligible (Ͻ3ϫ1010/cm2, about our sensitivity limit in
this study) numbers of paramagnetic centers. The low-field
stressing defect density is, within experimental error, the
same as that of the unstressed sample. In Fig. 2, we plot the
change in Pb density ͑⌬Pb͒ as a function of corona biasing
time for both high and lower values of corona bias on the
thick oxides.
In our study, we compare the response of oxide films to
both high and somewhat lower field corona bias for both
thick and thin film samples. Our results demonstrate quite
clearly that distinctions must be made between low- and
high-field corona charging. We find that low-field corona
charging appears to be relatively harmless, whereas high-
field corona-charging, as earlier studies8,9 have also shown,
is quite damaging. Our results clearly support both the use-
fulness of the corona biasing reliability tools and the validity
of earlier low-field corona ESR studies.
In stressing thin ͑75 Å͒ oxides we selected +6.5 MV/cm
for our high-field and +4 MV/cm as well as +1 MV/cm for
Our measurements utilized 75 and 634 Å SiO2 films
thermally grown in dry oxygen on relatively high
͑350–500 ⍀ cm͒ resistivity n-type silicon substrates with
(111) surface orientation. The (111) orientation was chosen
to simplify the ESR analysis. All wafers received a postdepo-
sition anneal in forming gas ͑N2/H2͒ at 400 °C for 30 min
and were cleaned using a standard SC1/SC2/HF process
FIG. 1. Pre- and post-corona bias ESR trace on thick ͑634 Å͒ oxides with
unstressed trace as reference. The slight asymmetry in the high-field trace is
a)Electronic mail: msd153@psu.edu
due to weak generation of E centers.
Ј
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
0003-6951/2004/85(10)/1844/2/$22.00 1844 © 2004 American Institute of Physics
155.247.166.234 On: Sun, 23 Nov 2014 01:52:45

Appl. Phys. Lett., Vol. 85, No. 10, 6 September 2004
Dautrich et al.
1845
defects, the somewhat lower field corona biasing
͑ഛ4 MV/cm͒ provides a relatively noninvasive means of
applying oxide bias. The results strongly indicate that the
much lower corona fields ͑ഛ2 MV/cm͒ utilized in reliability
tools and in the earlier low-field corona biasing ESR
studies3–7,11 had no damaging effect. More precisely, with
sensitivity threshold of about 3ϫ1010 paramagnetic
defects/cm2, we observe no defect generation in samples
stressed at low-field corona bias. Since this baseline is below
the sensitivity limit of the earlier low-field corona biasing
ESR studies3–7 and the same order of magnitude as the defect
density level in device quality interfaces, we conclude that
little or no relevant damage occurs to the Si/SiO2 system as
a result of this low-field corona biasing. (It should probably
be noted that our results may or may not be relevant to ex-
tremely thin oxides since our investigation involved oxides
no thinner than 75 Å).
Additionally we find that the overall response to corona
bias is essentially the same in both our thick and thin oxides:
high fields are damaging and low fields are not. Our results
also support earlier studies8,9 which showed that high electric
fields generated by corona ions are similar to high electric
fields generated by more conventional means, in that both
damage the oxides. In fact, as earlier studies have argued,8,9
high-field stressing via corona biasing may be at least very
roughly equivalent to high-field stressing via conventional
gate biasing. However, at lower fields, corona biasing, as
conventional biasing causes negligible damage. Thus a dis-
tinction must be made between high- and low-field corona
biasing.
FIG. 2. Comparison of high ͑+7 MV/cm͒ and lower ͑+4 MV/cm͒ corona
biasing time vs defect generation on thick ͑634 Å͒ oxides. Note: near neg-
ligible Pb center generation occurs in the lower corona bias samples.
our low-field corona biasing. As was the case for thicker
oxides, the high-field stressed sample exhibited generation of
very large densities ͑Ϸ2.1ϫ1012/cm2͒ of P and E centers
Ј
b
whereas the low-field stressed samples showed negligible
generation of centers (pre- and post-stressed samples both
exhibited Ϸ2ϫ1011 spins/cm2). Figure 3 depicts the change
in Pb density in thin samples as a function of stressing time.
Note that the oscillatory pattern is not experimental error and
indicates competition between the P and E centers. This is
Ј
b
consistent with the model proposed by Lenahan and
Conley,13–16 which involves the minimization of Gibbs free
energy of two systems of silicon dangling bonds with hydro-
gen passivation. The creation of oxide silicon dangling bond
defects, E centers, unpassivated by hydrogen, in the pres-
Ј
ence of a high density of Si/SiO2 interface silicon dangling
bonds, passivated by hydrogen, creates a situation in which
the system’s Gibbs free energy may be lowered by transfer-
ring hydrogen from one type of site (hydrogen passivated Pb
We find that whereas high-field corona bias induces large
densities of paramagnetic centers, lower-field corona biasing
generates virtually none. Our results thus support the validity
of the earlier low-field corona bias work. Our results also
indicate that commercial oxide characterization tools based
on the corona principle are not compromised by shortcom-
ings inherent in the technique. We find that corona biasing,
as does conventional gate biasing, causes negligible damage
at low fields, but generates numerous defects at high fields.
centers) to another (unpassivated oxide E centers).13–16 De-
Ј
pending upon the balance between the passivated E and P
Ј
b
centers, the reaction may be driven either way, toward more
passivated P centers or more passivated E centers.13–16
Ј
b
Additional wider magnetic field ESR scans of our high
field ͑6.5 MV/cm͒ thin oxide (not shown) also indicate sev-
eral defect centers including the generation of E , P , and
Ј
b
1KLA-Tencor (Milpitas, CA 95035) manufactures an in-line monitoring
and characterization tool called Quantox XP (http://www.kla-tencor.com/).
2M. Wilson, J. Lagowski, A. Savtchouk, L. Jastrzebski, and J. D’Amico,
ASTM Conference on Gate Dielectric Oxide Integrity, San Jose, CA,
1999.
possibly the nonbridging oxygen defect.
Our results clearly demonstrate that, whereas the higher
field corona bias ͑6.5 MV/cm͒ generates large numbers of
3P. M. Lenahan and P. V. Dressendorfer, J. Appl. Phys. 55, 3495 (1984).
4P. M. Lenahan and P. V. Dressendorfer, Appl. Phys. Lett. 41, 542 (1982).
5P. M. Lenahan and P. V. Dressendorfer, J. Appl. Phys. 54, 1457 (1983).
6Y. Y. Kim and P. M. Lenahan, J. Appl. Phys. 64, 3551 (1988).
7G. J. Gerardi, E. H. Poindexter, and P. J. Caplan, Appl. Phys. Lett. 49, 348
(1986).
8W. L. Warren and P. M. Lenahan, Appl. Phys. Lett. 49, 1296 (1986).
9W. L. Warren and P. M. Lenahan, IEEE Trans. Nucl. Sci. 34, 1355 (1987).
10A. Stesmans and V. V. Afanas’ev, Appl. Phys. Lett. 82, 2835 (2003).
11J. P. Campbell and P. M. Lenahan, Appl. Phys. Lett. 80, 1945 (2002).
12Z. A. Weinberg, W. C. Johnson, and M. A. Lampert, J. Appl. Phys. 47,
248 (1976).
13P. M. Lenahan, J. F. Conley, Jr., and B. D. Wallace, J. Appl. Phys. 81,
6822 (1997).
14J. F. Conley, Jr., P. M. Lenahan, B. D. Wallace, and P. Cole, IEEE Trans.
Nucl. Sci. 44, 1804 (1997).
FIG. 3. Comparison of high ͑+6.5 MV/cm͒, low ͑+4 MV/cm͒, and very
low ͑+1 MV/cm͒ corona biasing time vs defect generation on thin ͑75 Å͒
oxides. Note that as mentioned in the text, the oscillating behavior is almost
certainly not experimental error, but the result of a competition between the
15P. M. Lenahan and J. F. Conley, Jr., Appl. Phys. Lett. 71, 3126 (1997).
16P. M. Lenahan, J. J. Mele, J. F. Conley, Jr., R. K. Lowry, and D. Wood-
bury, IEEE Trans. Nucl. Sci. 46, 1534 (1999).
Pb and E sites for hydrogen.
Ј
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
155.247.166.234 On: Sun, 23 Nov 2014 01:52:45