Liquid-phase deposition of Al2O3 thin films on GaN

Article abstract of DOI:10.1149/1.2793700

Thin films of aluminum oxide (Al2 O3) on GaN substrates were grown by the process of low-temperature liquid-phase deposition (LPD), in which aluminum sulfate with crystallized water [Al2 (SO4) 3 18 H2 O] and sodium bicarbonate [NaHCO3] were used as the precursors. The pH value of the growth solution plays an important role in the deposition process. The best quality of the oxide thin film was obtained at the pH value of 3.80, while the growth rate was 35 nmh at the optimized concentration values of Al2 (SO4) 3 =0.0834 molL and NaHCO3 =0.211 molL and at the temperature of 30°C. The films were characterized by means of X-ray photoelectron spectroscopy, Auger electron spectroscopy, and atomic force microscopy. It was found that the leakage current density of 50 nm thin Al2 O3 oxide film was between 10-4 and 10-5 A cm2 at a negative electric field of 1 MVcm, with the breakdown electric field being greater than 10 MVcm. After annealing the oxide at 750°C for 30 min, the leakage current density was lowered to the value of 10-6 - 10-7 A cm2 at the negative electric field of 1 MVcm. The oxide-semiconductor interface state density as calculated from the capacitance-voltage curve was 3.89× 1011 cm-2 eV-1.

Full text of DOI:10.1149/1.2793700

Journal of The Electrochemical Society, 154 ͑12͒ H1041-H1046 ͑2007͒
H1041
0013-4651/2007/154͑12͒/H1041/6/$20.00 © The Electrochemical Society
Liquid-Phase Deposition of Al2O3 Thin Films on GaN
Sarbani Basu, Pramod K. Singh, Jian-Jiun Huang, and Yeong-Her Wangz
Department of Electrical Engineering, Institute of Microelectronics, Advanced Optoelectronic Technology
Center, National Cheng Kung University, Tainan 701, Taiwan
Thin films of aluminum oxide ͑Al2O3͒ on GaN substrates were grown by the process of low-temperature liquid-phase deposition
͑LPD͒, in which aluminum sulfate with crystallized water ͓Al2͑SO4͒3·18H2O͔ and sodium bicarbonate ͓NaHCO3͔ were used as
the precursors. The pH value of the growth solution plays an important role in the deposition process. The best quality of the oxide
thin film was obtained at the pH value of 3.80, while the growth rate was 35 nm/h at the optimized concentration values of
Al2͑SO4͒ = 0.0834 mol/L and NaHCO3 = 0.211 mol/L and at the temperature of 30°C. The films were characterized by means
3
of X-ray photoelectron spectroscopy, Auger electron spectroscopy, and atomic force microscopy. It was found that the leakage
current density of 50 nm thin Al2O3 oxide film was between 10−4 and 10−5 A/cm2 at a negative electric field of 1 MV/cm, with
the breakdown electric field being greater than 10 MV/cm. After annealing the oxide at 750°C for 30 min, the leakage current
density was lowered to the value of 10−6–10−7 A/cm2 at the negative electric field of 1 MV/cm. The oxide-semiconductor
interface state density as calculated from the capacitance-voltage curve was 3.89 ϫ 1011 cm−2 eV−1
© 2007 The Electrochemical Society. ͓DOI: 10.1149/1.2793700͔ All rights reserved.
.
Manuscript submitted January 16, 2007; revised manuscript received August 13, 2007. Available electronically October 22, 2007.
Experimental
Much attention has been focused on replacing SiO2 with Al2O3
as a dielectric insulation film for semiconductor device applications
due to the latter’s large bandgap ͑ϳ9 eV͒, higher dielectric constant
͑k = 10͒, high breakdown electric field ͑5–10 MV cm−1͒, good ther-
mal stability ͑amorphous condition up to 1000°C͒, chemical stability
against AlGaN ͑without interdiffusion and interaction of Si and Al͒,
and lower lattice mismatch to GaN. An amorphous Al2O3 film
shows a better barrier for alkali ions, impurities,1 and higher radia-
tion resistances. Because of these properties, ultrathin alumina films
of nanometer scale are widely used as high-k material to replace
SiO2 in microelectronic devices such as dynamic random access
This LPD deposition system9 contains a temperature-controlled
water bath that offers a uniform deposition temperature with 0.1°C
accuracy, a substrate holder, a Teflon beaker, a magnetic stirrer for
the high homogeneity of the growth solution, a pH meter, a Teflon
filter ͑0.1 ␮m͒, and an ultrasonic cleaner. The GaN epitaxial layer
͑2 ␮m͒ grown by MOCVD on a c-plane sapphire was used as the
substrate for the Al2O3 thin-film deposition. The substrates were
cleaned successively with acetone, methanol, and deionized ͑DI͒
water by sonicating each for 10 min and immersing them in hydrof-
luoric acid ͑HF͒ solution ͑1:1͒ for 1 min in order to remove the
organic contaminants and native oxides of GaN. Aluminum sulfate
͓Al2͑SO4͒3·18H2O͔ was placed in a Teflon beaker, and a small
amount of water was added to form a nearly saturated solution of
Al2͑SO4͒3 as a source liquid. Then, sodium bicarbonate fine powder
͑as a deposition rate controller͒ was added slowly to obtain a pH
value of 2.89 for the growth solution. After the addition of NaHCO3,
2
memories ͑DRAMs͒ and metal-oxide semiconductor field effect
transistors ͑MOSFETs͒ based on both Si and III-V compound
semiconductors.3 Applications of Al2O3 films are also being inves-
tigated in a variety of other areas including in optics, solid-state
electronics, and microelectromechanical systems ͑MEMS͒. In this
study, our aim was to deposit a thin Al2O3 insulating layer for GaN
metal oxide semiconductor high electron mobility transistor
͑MOSHEMT͒ device application to reduce the gate leakage current.
Many conventional methods for fabricating Al2O3 films are de-
scribed in different papers, including thermal oxidation,4 metallor-
ganic chemical vapor phase deposition ͑MOCVD͒,5 direct-current
reactive magnetron sputtering,6 photoluminescent alumina films by
pyrosol process,7 and atomic layer deposition ͑ALD͒.8
the hydrolysis of Al3+ particles and the concentration of Al͑OH͒
colloid particles increased, but no precipitation of the Al͑OH͒ par-
ticles was observed because the pH value was still low enough at
2.89.11 An aluminum sulfate solution is strongly acidic due to Al+3
ions, and it reacts with sodium bicarbonate powder, thereby gener-
ating carbon dioxide bubbles. As soon as the reaction was com-
pleted, DI water was immediately added in order to increase the pH
value to 3.80, which is the optimized pH value used for the deposi-
3
3
The growth of Al2O3 thin films as achieved by the above-
mentioned techniques involves high substrate temperatures
͑Ͼ750°C͒. This causes thermal strain and defect states in the semi-
conductor oxide interface, which in turn degrade device perfor-
mance. The present work is a new attempt to deposit amorphous
Al2O3 thin films on GaN substrates by means of the liquid phase
deposition ͑LPD͒ method at low temperatures in order to avoid ther-
mal strain and defects. This technique is found to incur low cost and
is reliable and efficient. In our previous work, we have successfully
demonstrated the deposition of a SiO2 thin film on a GaN substrate
by means of the LPD process at room temperature.9
tion of an alumina film. The final growth solution of the Al͑OH͒
3
solution was produced by the electrolytes of Al2͑SO4͒3 and Na2SO4.
The solution was diluted at the proper time so that no precipitation
occurred due to the crystallization of double sulfate ͓composed of
Na2SO4 and Al2͑SO4͒3͔. The growth solution was transparent, and
all the parameters were optimized with Al2͑SO4͒3 = 0.0834 mol/L
and NaHCO3 = 0.211 mol/L ͑final concentrations͒. The solution
was filtered through a 0.1 ␮m filter and was used for the film growth
on GaN substrates by means of the immersion method. In this
mechanism, Al͑OH͒ dissolved in acid when pH Ͻ 3.4 by generat-
3
ing Al+3 ions as well as in alkali when pH Ͼ 11 by generating
Recently, Sun et al.10 described the SiO2 and Al2O3 thin films
grown by LPD on GaAs and Si substrates; however, the aluminum–
oxygen ͑Al–O͒ bonding spectra of the grown film was not identified
by the X-ray photoelectron spectroscopy ͑XPS͒ study and the elec-
trical characteristics were not given. In the current research, we
therefore introduced a successful deposition of Al2O3 thin film on a
GaN substrate with material and electrical characterization.
Al͑OH͒ ions, but it remained undissolved when the pH value was
−
4
between 4 and 10.11 The precursor species Al2͑SO4͒3 and NaHCO3
were hydrolyzed and deposited on a clean GaN substrate. The Al2O3
films were prepared on the GaN substrate by the dehydration of
aluminum hydroxide.12 Postgrowth annealing was done to make the
films denser.
The chemical reaction of film growth can be described in the
following
⌬
z E-mail: yhw@eembox.ncku.edu.tw
Al2͑SO4͒3 + 6NaHCO3→2Al͑OH͒ + 3Na2SO4 + 6CO2↑
3
Downloaded on 2014-12-17 to IP 169.230.243.252 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

H1042
Journal of The Electrochemical Society, 154 ͑12͒ H1041-H1046 ͑2007͒
Figure 2. XPS spectra of a 70 nm thin Al2O3 film grown by LPD on a GaN
substrate. The film was annealed at 150°C for 2 h in an N2 environment
without any surface treatment. ͑a͒ Gaussian curve fitting of Al 2p peak and
͑b͒ Gaussian curve fitting of O 1s peak.
Figure 1. ͑a͒ Schematic of selective oxide growth and ͑b͒ film thickness vs
deposition time.
2Al͑OH͒ → Al2O3 + 3H2O
3
Al͑OH͒3 colloidal particles agglomerate together to form big colloi-
dal particles. Due to this, the solution became turbid after more than
3 h.
Results and Discussion
The GaN substrate has a transparency to He–Ne laser. Due to
that, the thickness of the oxide was determined by means of the
Dektak step-height measurement system through the selective depo-
sition of oxide as shown in Fig. 1a. The GaN substrate was spin
coated by photoresist, and the pattern was developed by the photo-
lithography process, then baked at 90°C for 10 min. These patterned
samples were immersed in a growth solution, and finally, the pho-
toresist was removed by acetone. Figure 1b shows the film thickness
of Al2O3 with deposition time at different growth temperatures. At
the initial stage the deposition rate is almost linear but within 2–3 h
Temperature was also a big factor in controlling film-growth
quality and deposition rate. At a higher temperature ͑i.e., 40°C͒, film
quality became poor, and the growth solution quickly became turbid.
For different oxide thicknesses, the refractive index of the deposited
samples varied from 1.50 to 1.65.
Figures 2a and b show the XPS of the Al 2p and O 1s core
levels, respectively. For XPS, we employed monochromatic Al K␣
͑1486.6 eV͒ radiation at a power of 300 W as the excitation source
and with a wide scan of 30 spectra. The take-off angle was set at 0°
relative to the surface normal. The thickness of the oxide film was
approximately 70 nm annealed at 150°C for 2 h without any surface
treatment. The spectrum marked by a solid square showed the ex-
perimental data. Only one peak of Al 2p at a binding energy of
75.16 eV was observed. Curve fitting was done using Gaussian
function ͑solid line͒ with a full width at half-maximum ͑fwhm͒ of
2.0 eV. This small fwhm value was closer to ␣-Al2O3 due to the
smallest Al–Al distance.13 In fact, aluminum has a wide range of
oxides ͑␣-Al2O3 and ␥-Al2O3͒, namely, hydroxides Al͑OH͒3, gibb-
site, bayerite, nordstrandite, oxyhydroxides ͑diaspore͒, and
␥-AlOOH ͑boehmite͒. Unfortunately, differentiating the Al 2p bind-
ing energy between different oxides, hydroxides, and oxyhydroxides
by XPS was difficult, but interpreting by X-alpha calculation from
it is nonlinear due to the nanoscale Al͑OH͒ colloidal particles ag-
3
glomerated together to form big colloidal particles, and the deposi-
tion rate is faster. It was observed that after more than 3 h, the
solution slowly started to become turbid due to the precipitation
phenomenon of Al͑OH͒ and the pH value of the solution was de-
3
creased from 3.80 to 3.67. The slight decrease in pH value was
caused by the consumption of NaHCO3 in the reaction. With the
addition of NaHCO3 to the Al2͑SO4͒3 solution, the hydrolysis of
Al3+ particles took place, and the concentration of Al͑OH͒ colloid
3
particles increased. Upon dilution of the above solution, the pH
value increased to 3.80. The solubility of Al͑OH͒3 was low near the
pH value of 3.80.11 With the progression of time the nanoscale
Downloaded on 2014-12-17 to IP 169.230.243.252 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 154 ͑12͒ H1041-H1046 ͑2007͒
valance band spectra13 was feasible. The Al 2p peak binding energy
H1043
Table I. The Al 2p, O 1s peak binding energy and fwhm of dif-
of metallic aluminum was 72.8 eV; however, no other peaks were
found within the range of 65–75 eV ͑Al2O3͒. This indicates that Al
metals were not present on the surface, as they would have resulted
in a shift of the Al 2p peak toward a lower binding energy. The O 1s
peak existed at 532.07 eV with a small fwhm of 2.1 eV; however,
no other peaks were observed in the higher and lower binding en-
ergy sides of the O 1s region. This is due to the absence of other
impurities and the impossibility of contamination in the oxide film.
All data of O 1s binding energy were easily identified as similar to
those of ␣ Al2O3.14 Besides, a symmetric XPS peak without shoul-
ders and a good curve fitting by Gaussian implied that the oxide was
of good chemical purity. The O/Al atomic ratio was 1.48, which was
calculated by the area sensitivity factor. The survey spectrum of the
surface did not show any other peaks of sodium, sulfur, gallium, and
ferent LPD-grown Al2O3 on GaN samples, in which the GaN sur-
faces were treated by H3PO4, NH4OH, and without any pretreat-
ment, followed by different annealing temperatures.
O 1s
Peak 1
B.E.a
fwhm
͑eV͒
Al 2p
Peak 1
B.E.
fwhm
͑eV͒
O 1s
Peak 2
B.E.
Al 2p
Peak 2
B.E.
Sample treatment
and annealing
temperature
Sample
A
͑eV͒
͑eV͒
H3PO4
As deposited
H3PO4
600°C, 30 min
H3PO4
531.46 527.61
1.9
531.66 527.78
1.9
532.79 528.24
2.4
75.06
1.7
75.17
1.7
76.06
1.8
75.16
2
76.16
76.47
77.53
–
B
C
D
E
1100°C, 30 min
nitride. Because the solubility of Al͑OH͒ at the condition ͑pH
3
No surface treatment 532.07
–
ϳ 3.80͒ of film growth was low, the growth of Al2O3 thin film on a
clean GaN substrate was favored. However, Na2SO4 was dissolved
in the solution and not deposited on the substrate.
150°C, 2 h
NH4OH/H2O͑1:1͒
600°C, 30 min
2.1
531.79
2.18
–
75.15
1.78
75.84
In order to investigate the effect of deposition rate and film
quality15 as well as to observe the peak binding-energy shifts by
means of XPS, five different samples ͑A–E͒ with different surface
pretreatments before oxide growth and with varying postannealing
temperatures were used in this work.
a Binding energy.
We used two different pretreatment solutions, namely, phos-
phoric acid ͑H3PO4͒ and ammonium hydroxide ͑NH4OH͒, as the
acidic and base medium, respectively. GaN is highly stable to most
chemical elements, so there is no suitable wet etchant for device
fabrication. Commercially available H3PO4 ͑85%͒ was used in the
original concentration. H3PO4 solution was used to remove the
nitrogen-depleted layers of GaN. The damage that occurred by high-
temperature annealing or the dry etching of the MOCVD of the GaN
epitaxial layer grown on a c-plane sapphire was also removed by
treatment with H3PO4 solution.16 The etching rate of GaN was con-
trolled by the hydration of the Ga ion. The ortho and meta phos-
phoric ions were the two attacking ions for GaN etching. The de-
tailed mechanism of GaN etching in H3PO4 ͑85%͒ can be found in
Ref. 17.
The surface of GaN exhibits Ga-rich phase, and Ga2O3 is the
dominant component for surface natural oxide.18 After NH4OH
treatment, two peaks of O–H bond19,20 and Ga2O3 bond21,22 ap-
peared, and a significant decrease in surface natural oxide was ob-
served. Thus, NH4OH pretreatment is effective to create a mixed
hydroxyl and oxide layer on the substrate surface, and the OH for-
mation is favorable15 for the growth of the oxide.
Samples A, B, and C were treated with H3PO4 ͑85%͒ at 80°C for
2 min. No treatment was given to sample D, while sample E was
treated with a NH4OH pH-controlled solution ͑pH 11.8, 25°C, 29%
NH4OH:H2O = 1:1͒. After the deposition of the oxide, the samples
were annealed at different temperatures in a nitrogen environment.
Sample A was an as-deposited oxide, sample B was thermally an-
nealed at 600°C for 30 min, sample C was thermally annealed at
1100°C for 30 min, sample D was annealed at 150°C for 2 h, and
sample E was annealed at 600°C for 30 min. All the data for oxide
signals of Al 2p and O 1s peak are summarized in Table I. For all
samples, the binding energy shifted toward the higher energy side
due to the variation in annealing temperatures and the nonstoichio-
metric nature of the deposited oxide. The difference in binding en-
ergy between the core levels was almost the same for each sample,
except for sample C due to a higher annealing temperature of
1100°C. Because of this, GaN remained thermally stable in the N2
atmosphere up to 900°C. At higher temperatures, significant decom-
position occurred and degraded the structural and morphological
properties of the film.
a lower binding energy of metallic aluminum ͑72.8 eV͒. Figures
4a-c show the O 1s peak binding energy of samples A, B, and C,
respectively. Furthermore, in samples A, B, and C, the O 1s peak 2
was found at the ranges of 527.61, 527.78, and 528.24 eV, respec-
tively, which could be due to an impurity like the PO−4 ion during the
surface treatment before deposition or to the nonstoichiometry of Ga
oxide. Significantly, we noted that the pretreatment of phosphoric
acid is commonly used in strong adhesive bonding with a GaN
substrate and is more resistant to attack by water vapor. Thus, we
concluded from these peak results of Al 2p and O 1s that the grown
film was entirely composed of amorphous Al2O3 film.14 Moreover,
we observed continuous binding-energy shifts among the different
samples, which could be due to the native defects encountered by
the change in relaxation energy of misplaced atoms and oxygen or to
the aluminum vacancies. For example, there were donor states as O
vacancies and acceptor states as Al vacancies. Another important
observation was that the core-level peaks were shifted in the same
way. The shift could be due to a change in the position of the Fermi
level in the forbidden bandgap.14 As the results show, the composi-
tion of oxide films may be attributed to Al2O3. However, the differ-
ent pretreatment solutions did not show any noticeable change in
growth rate and surface roughness.
Figure 5 shows the typical Auger electron spectroscopy ͑AES͒
depth profile data for 30 nm Al2O3 on a GaN substrate annealed at
450°C for 30 min in N2 atmosphere. The AES data revealed that the
composition of aluminum ͑Al͒ and oxygen ͑O͒ atoms in oxide film
was uniform from the surface to the oxide-semiconductor interface.
Small gallium ͑Ga͒ outdiffusion from the GaN substrate and Al
indiffusion into the GaN region during alumina thin-film deposition
was also observed. The films were nearly stoichiometric, with an
approximately constant composition of Al and O atoms from the
bulk to the interface. This result also confirms a good interface with
Al and O as well as the absence of peaks in the interface region.
Figure 6 displays the surface morphologies of oxides on GaN for
different annealing temperatures. The atomic force microscopy
͑AFM͒ data ͑the scan rate was 1.3 Hz and the set point was
0.158 V͒ shows that the root-mean-square ͑rms͒ surface roughness
of a 50 nm thick Al2O3 film was ͑height modulation in the 2
ϫ 2 ␮m2͒ only 1.025 nm after high-temperature annealing at
750°C, which was low as compared to SiO2 on a GaN substrate
͑5.2 nm͒.9 The surface morphology was improved after high-
temperature annealing. Therefore, this smooth-surface result pro-
vides potential for GaN MOS-HEMT device applications.
Figures 3a-c show the Al 2p peak binding energy shifts of
samples A, B, and C, respectively. The surface was treated by
H3PO4 solution, and different annealing temperatures were applied.
Again, we did not observe any shift in the Al 2p peak values toward
Downloaded on 2014-12-17 to IP 169.230.243.252 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

H1044
Journal of The Electrochemical Society, 154 ͑12͒ H1041-H1046 ͑2007͒
Figure 4. O 1s peak binding energy shifts; the surface was treated by H3PO4
solution following different annealing temperatures. ͑a͒ Sample A was as-
deposited, ͑b͒ sample B was annealed at 600°C for 30 min, and ͑c͒ sample C
was annealed at 1100°C for 30 min.
Figure 3. Al 2p peak binding energy shifts; the surface was treated by
H3PO4 solution following different annealing temperatures. ͑a͒ Sample A
was as-deposited, ͑b͒ sample B was annealed at 600°C for 30 min, and ͑c͒
sample C was annealed at 1100°C for 30 min.
with a carrier concentration of 1 ϫ 1018 cm−3 was used as the sub-
strate to measure the leakage current density. The procedures for
fabrication of the device were as follows. To form the ohmic con-
tacts, a Ti/Al layer was deposited by sputtering, followed by a 30 s
rapid thermal annealing at 900°C in nitrogen atmosphere. Then
Al2O3 oxide film was grown on the substrate. Finally, Al metal pads
were evaporated on the oxide as anodic plates with an area of
All these results indicate that the good physical and chemical
quality of Al2O3 films can be properly deposited on GaN substrates
by the LPD process.
For the investigation of electrical properties of the oxide film, a
device structure made of MOS diodes was employed. The Si-doped
n-type GaN epitaxial layer grown on a c-plane sapphire substrate
Downloaded on 2014-12-17 to IP 169.230.243.252 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 154 ͑12͒ H1041-H1046 ͑2007͒
H1045
Figure 5. Typical AES depth profiles of an aluminum oxide film deposited
on a GaN substrate.
1–7 ϫ 10−5 cm2. Figure 7a shows the typical leakage current den-
sity of LPD-grown Al2O3 thin film on GaN annealed at 150°C for
30 min. At a negative electric field of 1 MV/cm, the corresponding
leakage current densities ranged from 10−4 to 10−5 A/cm2. The
breakdown electric field was more than 10 MV/cm for a 50 nm
oxide thickness. Furthermore, the leakage current density could be
improved by a high-temperature annealing of oxide films. Figure 7b
shows the improvement of leakage current density with a higher
annealing temperature in N2 ambient. After annealing at 750°C,
the leakage current density was reduced to the order of
10−6–10−7 A/cm2 at a negative electric field of 1 MV/cm. This leak-
age current density of thin Al2O3 film on GaN was quite comparable
to or better than that of SiO2 on GaN9 and Al2O3 on Si by
sputtering.23 However, the result shows that the breakdown field was
higher. Electronic conduction may occur within the oxides via hop-
ping transport or tunneling process between trap sites, which may be
caused by unattached bonding ͑defects͒, porous structures ͑pin-
holes͒, or impurities. As a result, pinhole-free insulating oxides have
better electrical insulating properties such as lower leakage currents
and higher breakdown fields.24 The capacitance-voltage ͑C-V͒ char-
acteristics yielded the bulk and interface defect densities of the ox-
ide. Figure 8 shows the typical C-V characteristics measured by
4280 A at frequency 1 MHz, exhibiting the GaN surface-charge re-
gimes of electron accumulation, depletion, and inversion. At posi-
tive dc bias applied to the top Al gate terminal, the net capacitance
equaled the oxide capacitance Cox due to the accumulation of elec-
trons at the n-type GaN surface. In the case when moderate negative
voltage is applied to the top gate, a positive space-charge region is
induced in the n-type semiconductor. By implying a sufficiently
Figure 7. ͑a͒ The log I-V characteristics for the 50 nm thick Al2O3 on GaN
annealed at 150°C for 30 min. ͑b͒ The improvement of leakage current den-
sities with varying annealing temperatures.
Figure 6. ͑Color online͒ AFM ͑3D image͒ surface morphology of Al2O3 oxide deposited on a GaN substrate for different annealing temperatures in N2 ambient
for 30 min.
Downloaded on 2014-12-17 to IP 169.230.243.252 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

H1046
Journal of The Electrochemical Society, 154 ͑12͒ H1041-H1046 ͑2007͒
Conclusion
In summary, we have achieved a low-cost, more efficient, and
low-temperature ͑ϳ30°C͒ liquid-phase deposition of Al2O3 thin
films on GaN ͑deposition rate of 35 nm/h͒. The XPS study of LPD
Al2O3/GaN with and without H3PO4 and NH4OH pretreatments
were investigated in order to observe the shift in peak binding en-
ergy. Different samples did not show significant shifts in the peak
binding energy of Al 2p ͑75 eV͒ and O 1s ͑532 eV͒ core levels.
The good quality of alumina thin films was obtained by optimization
of Al2͑SO4͒ = 0.0834 mol/L and NaHCO3 = 0.211 mol/L concen-
3
trations. The refractive index and relative permittivity of oxide were
1.63 and 9.7, respectively. The various analytical results through
XPS, AES, and AFM showed that a reliable and good-quality alu-
minum oxide film could be obtained. Moreover, the results showed
that in terms of leakage current density, breakdown electric field,
and interface trap charge density, this LPD-Al2O3/GaN method pro-
vides a unique opportunity to make high-quality gate dielectrics for
GaN MOSFET applications.
Figure 8. The measured and ideal C-V characteristics at 1 MHz. The calcu-
lated interface trap densities were 3.89 ϫ 1011 cm−2 eV−1 for an oxide thick-
ness of 50 nm on GaN annealed at 150°C ͑30 min͒ and 5.59
ϫ 1011 cm−2 eV−1 for as-grown oxide.
Acknowledgments
This work was supported in part by the National Science Council
under contract no. NSC94-2215-E-006-001 and NSC95-2221-E-
428-MY3 by the Ministry of Education Program for Promoting Aca-
demic Excellence of Universities under grant no. A-91E-FA08-1-4,
and by the Foundation of Chen under the Jieh-Chen Scholarship of
Tainan, Taiwan.
large negative voltage to the gate of the MOS capacitor, an inversion
layer of holes was induced at the oxide-semiconductor interface.
The thin line and thick solid lines in Fig. 8 represent the experimen-
tal results of the as-grown and annealed films at 150°C for 30 min in
N2 ambient, respectively. The approximate curve of ideal C-V char-
acteristics based on theoretical calculation is also shown in broken
lines. According to the high-frequency ͑Terman’s͒ method,25 the dis-
tortion of the C-V curves is caused by a high charge density at the
interface. Obviously, there was a lower interface charge density for
the annealed oxide film than that for the as-grown film. The inter-
face charge density ͑Dit͒ and flatband voltage ͑VFB͒ can be deter-
National Cheng-Kung University assisted in meeting the publication
costs of this article.
References
1. R. A. Abbott and T. I. Kamins, Solid-State Electron., 13, 565 ͑1970͒.
2. G. D. Wilk, R. M. Wallace, and J. M. Anthony, J. Appl. Phys., 89, 5243 ͑2001͒.
3. K. P. Pande, V. K. R. Nair, and D. Gutierrez, J. Appl. Phys., 54, 5436 ͑1983͒.
4. J. Kolodzey, E. A. Chowdhury, T. N. Adam, G. Qui, I. Rau, J. O. Olowolafe, J. S.
Suehle, and Y. Chen, IEEE Trans. Electron Devices, 47, 121 ͑2000͒.
5. J. S. Kim, H. A. Marzouk, P. J. Reucroft, J. D. Robertson, and C. E. Hamrin, Jr.,
Appl. Phys. Lett., 62, 681 ͑1993͒.
6. M. Voigt and M. Sokolowski, Mater. Sci. Eng., B, 109, 99 ͑2004͒.
7. P. Tristant, Z. Ding, Q. B. Trang, V. H. Hidalgo, J. L. Jauberteau, J. Desmaison, and
C. Dong, Thin Solid Films, 390, 51 ͑2001͒.
8. H. C. Lin, P. D. Ye, and G. D. Wilk, Appl. Phys. Lett., 87, 182904 ͑2005͒.
9. H. R. Wu, K. W. Lee, T. B. Nian, D. W. Chou, J. J. Huang Wu, Y. H. Wang, M. P.
Huang, P. W. Sze, Y. K. Su, S. J. Chang, C. H. Ho, C. I. Chiang, Y. T. Chern, F. S.
Juang, T. C. Wen, W. I. Lee, and J. I. Chyi, Mater. Chem. Phys., 80, 329 ͑2003͒.
10. J. Sun, L. H. Hu, Z. Y. Wang, and G. T. Du, Mater. Sci. Forum, 475-479, 1725
͑2005͒.
−1
mined by employing the equation Dit = Cox/q ͓͑d¢s/dV͒ − 1͔
− CD/q, where Cox and CD are the oxide and depletion capacitance,
respectively. The surface potential ͑¢s͒ is the potential difference
across the space-charge layer. At ¢s = 0, the applying bias voltage
represents the VFB. A higher slope value is responsible for a lower
state density. According to the plot and calculated results, the VFB
was +2 and +0.2 V for the as-grown and annealed samples, respec-
tively. The calculated interface trap densities were 3.89
ϫ 1011 cm−2 eV−1 for an oxide thickness of 50 nm annealed at
150°C ͑30 min͒ and 5.59 ϫ 1011 cm−2 eV−1 for the as-grown oxide
film. Thus, after annealing, the average Dit near the midgap of
the LPD-grown oxide was reduced from 5.59 to 3.89
ϫ 1011 cm−2 eV−1. In addition, the C-V characteristics in the accu-
mulation region exhibited an almost completely horizontal curve,
which showed that the leakage current of the oxide was low enough
to accumulate majority carriers. According to the measured capaci-
tance at a dc bias voltage under accumulation, the calculated relative
11. D. F. Shriver, P. W. Atkins, and C. H. Langford, Inorganic Chemistry, Chap. 5, p.
164, Oxford University Press, New York ͑1990͒.
12. Z. D. Zivcovic and B. Dobovisek, J. Therm. Anal., 12, 207 ͑1977͒.
13. S. Thomas and P. M. A. Sherwood, Anal. Chem., 64, 2488 ͑1992͒.
14. W. M. Mullins and B. L. Averbach, Surf. Sci., 206, 52 ͑1988͒.
15. M. P. Houng, C. J. Huang, Y. H. Wang, N. F. Wang, and W. J. Chang, J. Appl.
Phys., 82, 5788 ͑1997͒.
16. B. J. Kim, J. W. Lee, H. S. Park, Y. Park, and T. I. Kim, J. Electron. Mater., 27, 5
͑1998͒.
17. A. Shintani and S. Minagawa, J. Electrochem. Soc., 123, 706 ͑1976͒.
18. T. Hashizume, S. Ootomo, S. Oyama, M. Konishi, and H. Hasegawa, J. Vac. Sci.
Technol. B, 19, 1675 ͑2001͒.
19. I. Waki, H. Fujioka, K. Ono, M. Miki, and A. Fukizawa, Jpn. J. Appl. Phys., Part
1, 39, 4451 ͑2000͒.
permittivity ͑␧ ͒ of oxide was 9.7 for the sample annealed at 450°C.
r
20. Y. J. Lin and C. T. Lee, Appl. Phys. Lett., 77, 3986 ͑2000͒.
21. G. Hollinger, S. Skheyta-Kabbani, and M. Gendry, Phys. Rev. B, 49, 11159 ͑1994͒.
22. N. J. Watkins, G. W. Wicks, and Y. Gao, Appl. Phys. Lett., 75, 2602 ͑1999͒.
23. J. Kolodzey, E. A. Choudhary, T. N. Adam, G. Qui, I. Rau, J. O. Olowolafe, J. S.
Suehle, and Y. Chen, IEEE Trans. Electron Devices, 47, 121 ͑2000͒.
24. Physics and Chemistry of III-V Compound Semiconductor Interfaces, C. W. Wilm-
sen, Editor, Chap. 3, Plenum Press, New York ͑1985͒.
25. E. H. Nicolian and J. R. Brews, MOS (Metal Oxide Semiconductor) Physics and
Technology, Chap. 8, John Wiley & Sons, Inc., New York ͑1982͒.
26. Handbook of Thin Film Technology, L. I. Maissel and R. Glang, Editors, p. 6.12,
McGraw-Hill, New York ͑1983͒.
To estimate the ͑␧ ͒, we applied the equation Cmeasured = ␧
r
r
␧ ͑A/tox͒, where A is the area of the anodic plate of the capacitors,
o
tox = 50 nm is the oxide thickness, and Cmeasured = 12 pF. The esti-
mated figure is close to the value of the dielectric constant ͑10.6͒ of
sapphire ͑␣-Al2O3͒.26
As these results show, the proposed method is likely to have the
potential to achieve a high gate breakdown voltage and low interface
trap charge density for GaN MOSHEMT device applications.
Downloaded on 2014-12-17 to IP 169.230.243.252 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).