Appl. Phys. A 77, 103 108 (2003)
Applied Physics A
Materials Science & Processing
DOI: 10.1007/s00339-003-2102-z
m.a. rana1, Stoichiometric and structural alterations
t. osipowicz1
in GaN thin films during annealling
h.w. choi2
m.b.h. breese1
1
Research Centre for Nuclear Microscopy, Department of Physics, National University of Singapore,
f. watt1
Singapore 117542, Singapore
2
s.j. chua2
Department of Electrical Engineering, Centre for Optoelectronics, National University of Singapore,
Singapore 119260, Singapore
plasma etching [11, 12], smooth biasing contacts and reduce
Received: 23 December 2002/Accepted: 16 January 2003
the roughness of the GaN-sapphire interface [13], remove
Published online: 28 March 2003 " © Springer-Verlag 2003
ion implantation damage [14], and clean surfaces [15, 16].
ABSTRACT Annealling experiments were performed on GaN
However, annealling above a certain temperature may cause
layers, grown on sapphire, over a wide range of tempera-
decomposition of GaN, especially near the surface, where ni-
ć%
tures (500 1100 C). Rutherford Backscattering Spectrometry
trogen and gallium evaporate, leaving vacancies, which may
(RBS) was performed in random and 0001 channelling ge-
result in the incorporation of oxygen. This incorporation of
ometries using 2 MeV protons and helium ions to determine
oxygen can cause n-type conductivity in initially p-type GaN,
the stoichiometric and structural alterations produced during
as discussed in theoretical studies by Mattila and Niemi-
annealling. We present here, for the first time, a comprehen-
nen [17], and Park and Chadi [18]. Strain can also be intro-
sive and quantitative analysis of the depth distribution of both
stoichiometric and structural changes in the near-surface re- duced during annealling due to the presence of oxygen and
gion (<" 750 nm) with a resolution of 50 nm for stoichiomet- water vapour in the ambient [19]. Pearton et al. [20] have
ric and 20 nm for structural changes. No decomposition was presented results of diffusion of oxygen into GaN during an-
ć%
ć%
measured for temperatures up to 800 C. Decomposition in the
nealling in the lower temperature range of 500 900 C under
near-surface region increased rapidly with further increases in
different experimental conditions to ours.
temperature, resulting in a near-amorphous region (500 nm) for
A number of authors [21 26] have studied the thermal
ć%
annealling at 1100 C. We describe the range of annealling
stability of GaN and related issues using experiment and
conditions under which negligible stoichiometric and struc-
simulation. In early studies, Lin et al. [21] studied the pho-
tural changes are observed. Our nanoscale resolution results
toluminescence line intensity of GaN samples annealled at
are useful for the fabrication and operation of conventional and
ć%
temperatures up to 900 C and found that GaN was degraded
nanoscale optoelectronic and high-temperature devices.
ć%
at 900 C. Vartuli et al. [22] studied the surface morphology
PACS 68.35.Dv; 61.85.+p; 61.72.Ji
and electrical conductivity of annealled GaN samples. They
also studied the surface stoichiometry of annealled GaN sam-
ples up to a depth of 30 nm. We have observed that lattice
1 Introduction alterations end at a considerably larger depth than 30 nm for
ć%
annealling above 900 C. Ambacher et al. [23] have reported
Gallium nitride has a wide range of applications in
results on the thermal stability of Group III nitrides including
optoelectronics due to its wide direct band gap and the possi-
GaN. They determined thermally induced hydrogen, hydro-
bility of band gap engineering [1 5]. It is one of the best semi-
carbon, and nitrogen-hydrogen effusion from thin films of
conductor materials for high-temperature/high-power elec-
these nitrides. Onozu et al. [24] used molecular dynamics cal-
tronic devices due to its excellent electronic and thermal prop-
culations to study the thermal stability of GaN thin layers.
erties, high breakdown voltage, chemical inertness and the
Kuball et al. [25] measured the effect of annealling on the free
possibility of fabrication of both unipolar and bipolar de-
carrier density in ion implanted GaN using UV Raman scat-
vices with low parasitics [6]. In addition to these conven-
tering, while Saarinen et al. [26] studied the thermal stability
tional applications, GaN and its alloys are now also of interest
of Ga vacancies produced during GaN growth. Despite a con-
due to their use in optoelectronic nanoscale devices, such as
siderable number of studies, a comprehensive investigation of
nanowires and nanorods [7, 8].
the depth distribution of stoichiometric and structural changes
High-temperature annealling of GaN is an important step
during annealling is missing.
in producing blue light emitters and high-temperature de-
Quantitative depth profiling of crystalline lattice disorder
vices. This process is used to activate Mg and Ca for p-type
and impurities are of great importance in semiconductor de-
doping of GaN [9, 10], reconstruct a stoichiometric near-
vice characterization as their performance depends strongly
surface region and improve the optical properties of GaN after
on the control of defects and impurities [27, 28]. A number
of authors [29 34] have used RBS/channelling to character-
Fax: [phone], E-mail: [e-mail] ize defects and impurities in crystals, especially in semicon-
104 Applied Physics A  Materials Science & Processing
ductors. Here we have determined the depth distribution of
lattice disorder and stoichiometric changes in GaN samples
annealled over a wide range of temperatures using RBS in ran-
dom and channelling geometries. The depth distributions of
defects have been determined assuming the displacement of
gallium atoms after evaporation of nitrogen and gallium.
2 Experimental
Annealling experiments were carried out on GaN
layers (3-µm-thick) which were grown on sapphire through
Metal Organic Chemical Vapour Deposition (MOCVD).
Samples were annealled at temperatures between 500 and
ć%
1100 C for a time interval of 60 s, with additional ramp
up and down times of 20 s. Annealling was carried out in
a nitrogen ambient and samples were kept uncapped dur-
ing annealling to study the process of decomposition. The
ć%
sample temperature was maintained within Ä…5 C from the
set temperature during annealling. Channelling and random
backscattering measurements were performed using 2MeV
FIGURE 1 Helium ion random backscattering spectra from GaN samples
helium ion and proton beams. Ion channelling measurements
annealled at different temperatures for 60 s. Annealling experiments were
ć%
also carried out at 500, 600, 700, 800, and 900 C but spectra are not shown
were carried out along the 0001 axis, and backscattered par-
to avoid overlapping. The depth scale corresponds to that of gallium atoms
ticles were measured using a semiconductor surface barrier
detector located at a scattering angle of 160ć%. Charge integra-
tion of the incident beam was achieved by applying a -500 V
bias voltage in front of the samples to suppress secondary
electron emission. The total beam fluence used for each meas-
urement was 5 10 µC.
3Results
3.1 Random RBS measurements
3.1.1 Gallium measurements. Figure 1 shows random back-
scattering spectra from samples annealled at different tem-
peratures using 2MeV He ions. Vertical arrows show the
positions along the energy scale where signals from gallium,
oxygen and nitrogen atoms at the surface are expected to ap-
pear. Stoichiometric changes are observed in the near-surface
region during annealling. Loss of gallium atoms from the
near-surface region of annealled GaN is manifested as a de-
crease in the backscattering signal from gallium atoms. The
presence of oxygen in the near-surface region of the GaN sam-
ples annealled at high-temperatures can be seen in Fig. 1, but
FIGURE 2 Measured gallium content of the near-surface region (<" 750 nm)
the ability to detect oxygen with He ions is poor due to the of annealled GaN samples, as a function of temperature, determined from the
helium ion backscattering spectra in Fig. 1. The nitrogen and oxygen content
small scattering cross-section of lighter target atoms. More
were determined from the backscattering spectra shown in Fig. 4. Lines are
precise measurements of oxygen and nitrogen profiles using
drawn to guide the eye
2MeVprotons are presented in the next section. Helium ions
were used for the gallium measurements as they provide better
depth resolution compared with protons. in Fig. 2 as a function of annealling temperature. The percent-
With increasing temperature, nitrogen and gallium can age of gallium atoms in the near-surface region of annealled
diffuse from deeper regions of the lattice to the surface. This GaN samples was also determined as a function of depth using
decomposition of GaN initiates two processes. One is the dis- the same measurements of as-grown and annealled GaN sam-
placement of atoms to new equilibrium lattice positions, and ples and is shown in Fig. 3. These gallium results are plotted
ć%
the other is the diffusion of oxygen from the ambient into the only for temperatures of 1000, 1050, and 1100 C. For lower
GaN to fill the vacancies produced by the evaporation of nitro- annealling temperatures, no significant change in the gallium
gen and gallium. The backscattering spectra were quantified signal was observed in RBS random measurements.
using the simulation code SIMNRA [35] for different an-
nealling temperatures. The percentage of gallium in the near- 3.1.2 Nitrogen and oxygen measurements. This section con-
surface region (<" 750 nm) after annealling, determined by fit- tains results of RBS random measurements of annealled
ting the helium ion backscattering spectra in Fig. 1, is shown GaN samples using 2MeV protons. Figure 4 shows random
RANA et al. Stoichiometric and structural alterations in GaN thin films during annealling 105
FIGURE 5 Experimental and simulated (SIMNRA [35]) random backscat-
FIGURE 3 Percentage of gallium atoms in annealled GaN, as a function of
ć%
tering proton spectra from as-grown and 1100 C annealled GaN samples.
depth, determined from the helium backscattering spectra in Fig. 1. Lines are
ć%
Solid and broken lines show simulated spectra for as-grown and 1100 C
drawn to guide the eye
samples, respectively.
FIGURE 4 Proton backscattering spectra from GaN samples annealled at
FIGURE 6 Widths of the depleted gallium (WGa) and nitrogen (WN) re-
different temperatures. The spectral region between 1600 and 1800 keV has
gions and oxygen-incorporated (WO) regions for annealled GaN as functions
been omitted to highlight the variation of the signals for Ga, O and N. Vertical
of the annealling temperature. Lines are drawn to guide the eye
arrows show the positions at which the signals from the corresponding elem-
ents (Ga, O or N) are expected to appear. Lines are drawn to guide the eye
at a progressively faster rate. Oxygen is gained by the lattice
backscattering measurements of the same as-grown and an- at approximately the same rate. In our experiment, oxygen is
nealled GaN samples. Only selected spectra are shown for present in the ambient, which is nitrogen containing oxygen as
the purpose of clarity. The effects of evaporation of nitro- an impurity.
gen and incorporation of oxygen during annealling are now The near-surface region becomes depleted of nitrogen and
clearly separated due to the higher scattering cross-section gallium during annealling due to evaporation. Oxygen from
of protons compared with helium ions for lighter elements. the ambient can diffuse into a portion of the disrupted region
Evaporation of nitrogen and incorporation of oxygen were near to the surface where depletion is greatest. The widths of
determined quantitatively by simulating these backscattering the regions depleted of gallium and nitrogen and the oxygen-
spectra. Two representative simulated spectra for as-grown incorporated regions, as determined from the 2MeV proton
ć%
and 1100 C annealled GaN samples are shown in Fig. 5. beam RBS random spectra, are shown in Fig. 6. At tempera-
ć%
The measured percentage of nitrogen and oxygen in the tures of > 1000 C, the width of the depleted region extends
near-surface region of annealled GaN samples using 2MeV at least 100 nm beyond the oxygen-incorporated region. Figs.
protons is shown in Fig. 2. It can be seen that nitrogen starts 7 and 8 show the atomic percentages of nitrogen and oxy-
to evaporate at a lower temperature than gallium and is lost gen atoms present in annealled GaN samples as functions of
106 Applied Physics A  Materials Science & Processing
FIGURE 9 Channelling backscattering spectra from 0001 axially aligned
FIGURE 7 Percentage of nitrogen present in the near-surface region of
GaN samples, which were annealled at different temperatures. The horizontal
GaN, as a function of depth, determined from proton backscattering meas-
arrow shows the region used for the determination of Çmin. Vertical arrows
urements. Lines are drawn to guide the eye
show the positions where backscattering signals from oxygen and nitrogen at
the surface appear. The depth scale corresponds to that of gallium atoms in
random backscattering
FIGURE 8 Percentage oxygen gain in the near-surface region of annealled
FIGURE 10 Minimum channelling yield, Çmin, for annealled GaN as a func-
GaN, as a function of depth, determined from proton backscattering measure-
tion of temperature. The line is drawn to guide the eye
ments. Lines are drawn to guide the eye
ć%
depth. These ratios were determined from the comparison of decomposition of the lattice at temperatures above <" 900 C.
the random spectra of the as-grown and annealled samples. The major cause is the displacement of atoms after the evap-
oration of nitrogen and gallium from their lattice sites, which
was discussed in the previous section. Channelling spectra
3.2 Channelling measurements
were also collected for annealling temperatures of 500, 600,
ć%
The random backscattering results using 2MeVhe- 700, and 800 C but are not shown in Fig. 9 to avoid overlap-
lium ion and proton beams show variations in the stoichio- ping of spectra.
metric depth distribution with annealling temperature. We Disruption and oxygen incorporation in the near-surface
now describe 2MeV helium ion channelling results, which region can be seen as increasing with temperature in Fig. 9.
show variations in crystalline quality as a function of depth. The signal from oxygen at the surface is more pronounced in
Figure 9 shows 0001 axially channelled RBS spectra for the channelling helium ion spectra due to reduced background
the same GaN samples. For annealling at temperatures up to counts compared to helium random spectra. Figure 10 shows
ć%
800 C, dechannelling within the GaN is negligible. Dechan- values of the minimum channelling yield, Çmin, as a function
nelling increases rapidly above this temperature, and is man- of annealling temperature. Figure 11 shows the percentage
ifested as an increasing backscattering signal from a thicker of displaced gallium atoms, as a function of depth, follow-
region close to the wafer surface. This indicates significant ing evaporation of nitrogen and gallium during isochronal
RANA et al. Stoichiometric and structural alterations in GaN thin films during annealling 107
of gallium. Incorporation of oxygen is substantial and most
gallium atoms are displaced from their lattice positions, cre-
ating an almost amorphous layer in the near-surface region.
The value of Çmin increases rapidly in this region and becomes
more than 50%.
ć%
We conclude that 700 800 C is the critical temperature
range for GaN for which lattice disruption starts to occur. This
ć%
increases slowly up to 900 C and then more rapidly with fur-
ther increases in temperature. These temperatures are quite
ć%
low compared to the melting point of GaN (> 1700 C). Afew
initial studies [36 38] provide information on the thermal
decomposition of GaN. Johnson et al. [36] have suggested
the evaporation of monomeric gaseous GaN occurs, whereas
Sim and Margrave [37] have suggested the evaporation of
polymeric gaseous GaN without decomposition takes place
ć%
at temperatures higher than 800 C. Further details are given
by Munir and Searcy [38] and the references therein and the
related theory is provided by Langmuir [39] and Hirth and
Pound [40]. There are two other possible routes by which GaN
FIGURE 11 Percentage of displaced gallium atoms out of total gallium
may decompose into gallium and nitrogen. In the first, gallium
present in annealled GaN samples, as a function of depth, determined from
channelling and random measurements. Lines are drawn to guide the eye
and nitrogen can be separate in gaseous form while in the sec-
ond, gallium can be in a liquid state [38]. In our results, we
annealling, based on the channelling measurements in Fig. 9. clearly observe that evaporation of nitrogen occurs at a con-
The number of displaced atoms was calculated by compar- siderably higher rate than that of gallium, which shows that
ing the backscattering gallium signal in the channelling and the evaporation of monomeric and polymeric gaseous GaN
random spectra [30] at each annealling temperature, equat- can only account for a fraction. Evaporation mainly occurs
ing the dechannelling signal to the fraction of gallium atoms after decomposition of GaN into gallium and nitrogen, sug-
displaced from their lattice sites. For temperatures beyond gesting the formation of gallium atoms in a liquid state, which
ć%
1000 C, the percentage of displaced gallium atoms reaches causes dechannelling, as shown in Fig. 11.
70%. Together, Figs. 3 and 11 give a complete picture of the
disruption of the gallium lattice. At a particular tempera-
ture, a small percentage of gallium atoms evaporates and
4 Interpretation and discussion
a large fraction of those remaining are displaced from their
In the previous section, the results of 2MeV He lattice sites. Thus, we may conclude that gallium atoms are
and proton beam experiments on as-grown and annealled GaN first displaced, then a fraction of them evaporate. Evapora-
were presented. Evaporation of gallium and nitrogen and the tion of nitrogen is found to be the dominant process in high-
subsequent incorporation of oxygen after the decomposition temperature annealling, as most nitrogen atoms are evapo-
of GaN can be seen clearly for samples annealled at high rated from the near-surface region during high-temperature
temperatures. The incorporation of oxygen in GaN is import- annealling. Similar profiles were measured for nitrogen evap-
ant because it can alter its doping level. Thus, if GaN-based oration and oxygen incorporation, suggesting that oxygen
transistors, in high power devices, are briefly exposed to tem- takes vacant nitrogen sites in the GaN lattice.
ć%
peratures higher than 900 C, their electronic properties may It is well known that hydrogen enhances the incorporation
change quite considerably. of a Mg-dopant in GaN due to the formation of Mg-H com-
In Figs. 2, 6, and 10, we may divide the range of annealling plexes, but it must be removed by post-growth annealling [41,
temperatures covered in our experiments into three regions. 42]. The presence of oxygen during low temperature an-
ć%
The first region includes temperatures up to 700 C, the sec- nealling helps in the removal of hydrogen, which is believed
ć%
ond temperatures from 700 to 900 C, and the third tempera- to be responsible for low p-type conductivities of GaN [43].
ć%
tures between 900 and 1100 C. In the first region, decompos- The presence of oxygen in the ambient during low tempera-
ć%
ition and the resulting evaporation and displacement of atoms ture annealling (up to 600 C) is useful for the activation of
from lattice positions is negligible and GaN is found to be Mg-doped GaN [43, 44]. Our results show that decomposition
thermally stable with no changes observed in its crystalline of GaN and changes in its stoichiometry are negligible during
ć%
quality. The value of Çmin is the same as that of as-grown annealling at temperatures lower than 700 C, at least for time
GaN. In the second region, there is measurable decompos- intervals of 60 s.
ition, resulting primarily in the evaporation of nitrogen and
displacement of gallium atoms from their lattice sites due to
5Conclusion
vacancies created after decomposition. The value of Çmin in-
creases slightly and a small amount of oxygen is incorporated. This study gives, for the first time, the width
In the third region, the decomposition is considerable, result- of the near-surface region altered during annealling over
ć%
ing in the evaporation of both gallium and nitrogen, although a wide range of temperatures (500 1100 C), along with the
the rate of evaporation of nitrogen is much higher than that nanoscale depth distribution of stoichiometric and structural
108 Applied Physics A  Materials Science & Processing
17 T. Mattila, R.M. Nieminen: Phys. Rev. B 54, 166 676 (1996)
changes in the altered region of annealled GaN samples.
18 C.H. Park, D.J. Chadi: Phys. Rev. B 55, 12 995 (1997)
The decomposition of GaN is found to be negligible dur-
19 J.M. Hayes, K. Kuball, A. Bell, I. Harrison, D. Korakakis, C.T. Foxon:
ć%
ing annealling up to 800 C. This range of temperatures is
Appl. Phys. Lett. 75, 2097 (1999)
therefore safe for the processing of GaN and related materials
20 S.J. Pearton, H. Cho, J.R. LaRoche, F. Ren, R.G. Wilson, J.W. Lee:
J. Appl. Phys. 75, 2939 (1999)
and operation of high-temperature devices of the same ma-
ć% 21 M.E. Lin, B.N. Sverdlov, H. Morkoc: Appl. Phys. Lett. 63, 3625 (1993)
terials. Thermal annealling of GaN beyond 1000 C causes
22 C.B. Vartuli, S.J. Pearton, C.R. Abernathy, J.D. MacKenzie, E.S. Lam-
considerable decomposition in the near-surface region fol-
bers, J.C. Zopler: J. Vac. Sci. Technol. B 14, 3523 (1996)
lowed by evaporation of nitrogen and gallium, which results 23 O. Ambacher, M.S. Brandt, R. Dimitrov, T. Metzger, M. Stutzmann,
R.A. Fischer, A. Miehr, A.M. Bergamier, G. Dollinger: J. Vac. Sci. Tech-
in the incorporation of oxygen from the ambient. Tempera-
ć% nol. B 14, 3532 (1996)
tures of more than 1000 C during fabrication and operation
24 T. Onozu, R. Miura, S. Takami, M. Kubo, A. Miyamoto, Y. lyechika,
of GaN-based devices will change their electronic and optical
T. Meda: Jpn. J. Appl. Phys. (Part 1) 39, 4400 (2000)
properties. 25 M. Kuball, J.M. Hayes, T. Suski, J. Jun, M. Leszczynski, J. Domagala,
H.H. Tan, J.S. Williams, C. Jagdish: J. Appl. Phys. 87, 2736 (2000)
26 K. Saarinen, T. Suski, I. Grzegory, D.C. Look: Phys. Rev. B 64, 233 201
REFERENCES (2001)
27 H.J. Queisser, E.E. Haller: Science 281, 945 (1998)
28 T. Wichert, M. Deicher: Nucl. Phys. A 693, 327 (2001)
1 O. Manasreh: III-Nitride Semiconductors: Electrical, Structural and
29 S.O. Kucheyev, J.S. Williams, C. Jagadish, J. Zou, G. Li: Phys. Rev. B
Defect Properties (Elsevier Science, Amsterdam, 2000)
62, 7510 (2000)
2 X. Duan, C.M. Lieber: J. Am. Chem. Soc. 122, 188 (2000)
30 L.C. Feldman, J.W. Mayer, S.T. Picraux: Materials Analysis by Ion
3 S. Nakamura: Science 281, 956 (1998)
Channeling: Submicron Crystallography (Academic Press, New York,
4 F.A. Ponce, D.P. Bour: Nature 386, 35 (1997)
5 S. Nakamuray, G. Fasol: The Blue Laser Diode: GaN Based Light Emit- 1982)
31 M.B.H. Breese, D.N. Jemieson, P.J.C. King: Materials Analysis using
ters and Lasers (Springer-Verlag, Berlin, Heidelberg, New York, 1997)
a Nuclear Microprobe (John Wiley & Sons, Inc., New York, 1996)
6 M.S. Shur: Mater. Res. Soc. Symp. Proc. 483, 15 (1998)
32 E. BÅ‚gh: Can. J. Phys. 46, 653 (1968)
7 Y. Huang, X. Duan, C.M. Lieber: Nano Lett. 2, 101 (2002)
33 W. Jiang, W.J. Weber, S. Thevuthasan: J. Appl. Phys. 87, 7671 (2000)
8 W. Han, S. Fan, Q. Li, Y. Hu: Science 277, 1287 (1997)
34 J.R. Tesmer, M. Nastasi: Handbook of Modern Ion Beam Materials An-
9 C.F. Lin, H.C. Cheng, C.C. Chang, G.C. Chi: J. Appl. Phys. 88, 6515
alysis (Material Research Society, USA, 1995)
(2000)
35 M. Mayer: SIMNRA Users Guide, Report IPP 9/113 (Max-Planck-
10 J.C. Zolper, R.G. Wilson, S.J. Pearton, R.A. Stall: Appl. Phys. Lett. 68,
Institut für Plasmaphysik, Garching, Germany, 1997-2001)
1945 (1996)
36 W.C. Johnson, J.B. Parsons, M.C. Crew: J. Phys. Chem. 36, 2651 (1932)
11 H.W. Choi, S.J. Chua, A. Raman, J.S. Pan, A.T.S. Wee: Appl. Phys. Lett.
37 R.J. Sim, J.L. Margrave: J. Phys. Chem. 60, 810 (1956)
77, 1795 (2000)
38 Z.A. Munir, A.W. Searcy: J. Chem. Phys. 42, 4223 (1965)
12 S. Tripathy, S.J. Chua, A. Raman, E.K. Sia, J.S. Pan, R. Lim, G. Yu,
39 I. Langmuir: Phys. Rev. 2, 329 (1913)
Z.X. Shen: J. Appl. Phys. 91, 3398 (2002)
40 J.P. Hirth, G.M. Pound: J. Chem. Phys. 26, 1216 (1957)
13 M.W. Cole, F. Ren, S.J. Pearton: Appl. Phys. Lett. 71, 3004 (1997)
41 F.A. Reboredo, S.T. Pantelides: Phys. Rev. Lett. 82, 1887 (1999)
14 H.H. Tan, S.J. Williams, J. Zou, D.J.H. Cockayne, S.J. Pearton,
42 J. Neugebauer, C. Van de Walle: Phys. Rev. Lett. 75, 4452 (1995)
J.C. Zolper, R.A. Stall: Appl. Phys. Lett. 72, 1190 (1998)
43 C.H. Kuo, S.J. Chang, Y.K. Su, J.F. Chen, L.W. Wu, J.K. Sheu,
15 L.L. Smith, S.W. King, R.J. Nemanich, R.F. Davis: J. Elect. Mater. 25,
C.H. Chen, G.C. Chi: IEEEElect. Dev. Lett. 23, 240 (2002)
805 (1996)
44 C.H. Kuo, S.J. Chang, Y.K. Su, J.F. Chen, L.W. Wu, J.K. Sheu,
16 K. Wang, D. Pavlidis, J. Cao: J. Elect. Mater. 26, 1 (1967)