J. Kim1, B. Gila2, R. Mehandru1,
J.W. Johnson1, J. H. Shin3, K.P. Lee2, B. Luo1,
A. Onstine2, C. R. Abernathy2, S.J. Pearton2,
and F.Ren1
1. Department of Chemical Engineering, University of Florida, Gainesville, FL 32611, USA.
2. Department of Materials Science and Engineering, University of Florida, Gainesville,
FL 32611, USA.
3. Samsung Electronics, Kyungki-Do, Korea.
GaN metal oxide
semiconductor diodes were demonstrated utilizing MgO as the gate oxide. MgO was
grown at 100 °C on the
MOCVD grown n-GaN in a molecular beam epitaxy system using a Mg elemental
source and an electron cyclotron resonance oxygen plasma. H3PO4 based
wet-chemical etchant was used to remove MgO to expose the underlying n-GaN for
ohmic metal deposition. Electron
deposited Ti/Al/Pt/Au and Pt/Au were utilized as ohmic and gate metallization,
respectively. An interface trap density of low-to-mid 1011 eV-1cm-2
was obtained from temperature conductance-voltage measurements. The Terman method was also
used to estimate the interface trap density and a slight lower number was
obtained as compared to the conductance method. Results from elevated temperature (up to 300°C)
conductance measurements showed an interface state density roughly three times
higher(6x1011 eV –1 cm-2 ) than at 25oC.
GaN based light
emitting and laser diodes have been receiving a lot of attention for the past
few years and those diodes are now commercially available[1-2]. GaN based
rectifiers and field effect transistors (FET) are also very attractive
for high temperature, high voltage, high power, and high frequency
applications. In order to reduce the gate leakage current and increase device
breakdown voltage,the metal oxide semiconductor (MOS) structure is preferred,
however, unlike in Si technology, GaN does not have high quality native
oxides. Gate dielectrics of AlN, Ga2O3(Gd2O3),
SiO2, and Si3N4 have all been previously
reported[3 –9]. Various deposition
techniques such as plasma enhanced chemical vapor deposition (PECVD), jet vapor
deposition, molecular beam epitaxy (MBE), metal organic chemical vapor
deposition (MOCVD), and electron beam deposition were used for those oxide
growths. Recently the bixbyite oxide Gd2O3, grown via gas
source molecular beam epitaxy on n-GaN, has been shown to provide a interface
state density of mid 1011 eV-1cm-2.However,
while Gd2O3 has shown promising results, its rather small
bandgap, 5.3eV, and large lattice mismatch to GaN, ~20%, leave substantial room
for improvement[10}. MgO is an attractive alternative as it has the same
crystal structure as Gd2O3, but a larger bandgap, 8eV,
and a smaller lattice mismatch to GaN, ~6.5%.
In this study, MgO
was grown on GaN using a Mg elemental source and an electron cyclotron
resonance (ECR) oxygen plasma in a MBE system.
To measure interface trap density,the
two methods,the Terman and the conductance technique were used to estimate the
interface trap density. Due to slow
minority-carrier generation for the wide energy gap material,the conductance
was also measured at high temperature to enhance the minority carrier
generation.
Oxide growth was performed in a RIBER 2300 MBE equipped with a reflection high-energy electron diffraction (RHEED) system. Substrates employed were MOCVD grown n-GaN on sapphire (0001). Oxygen was supplied from a Wavemat MPDR 610 electron cyclotron resonance (ECR) plasma source (2.54 GHz) with 200 watts forward power at 1x10-4 Torr of the oxygen pressure. A standard effusion cell operating at 1130°-1170°C was used for the evaporation of the magnesium. The substrate temperature was kept at 600 °C and measured using a backside thermocouple that was calibrated using the melting points of InSb and GaSb.
Prior to the oxide growth, the GaN substrate preparation consisted of a wet chemical etch of HCl:H2O (1:1) for 3 minutes, a DI rinse, a UV-ozone exposure for 25 minutes in a UV Cleaner model 42-220, a dip in buffered oxide etch (6:1, ammonium fluoride : hydrofluoric acid) for 5 minutes and a final DI rinse. The substrates were then indium mounted to molybdenum blocks and loaded into the MBE. At room temperature the surface of the substrates was polycrystalline according to the RHEED patterns. Upon heating the GaN to 700°C, a streaky (1×3) pattern appeared. This was the starting surface for all the films grown in this study.
The diode fabrication started with ohmic contact formation by removing MgO and depositing Ti/Al/Pt/Au. AZ-1045 and H3PO4 solution were used for the etch mask and oxide removal, respectively. A standard lift-off process was employed for the Pt/Au based gate deposition. Figure 1 illustrates a cross-section view of the diode.

Figure 1 Cross-sectional view of a GaN diode with MgO as the gate oxide.
Capacitance and conductance measurements were performed with an HP 4284 precision LCR meter. A hot chuck was used for high temperature conductance measurements and the temperature range was from room temperature to 300 °C.
Figure 2 show the I-V
characteristics at 25o C from a MgO/GaN diode with 100 nm MgO gate
dielectric. The reverse breakdown and
forward turn-on voltage are >40 V and >10 V, respectively, where these
parameters are defined as the values at 100 nA or 5 mA/cm2,respectively). A forward breakdown field of 1.2 MV/cm for
MgO was calculated from this data.

Figure 2. Diode characterisitics of an 100 nm thick MgO/GaN diode.
The measured C-V curve at a frequency of 1 MHz and sweep-rate of 100 mV/s, as illustrated in Figure 3, showed a clear deep depletion behavior for negative bias voltage and no measurable hysterisis was observed. The deep depletion feature with no inversion capacitance characteristics is typical of wide-gap semiconductor MIS structures due to the slow generation rate of the minority carrier at room temperature. Using the relation of Cox = eoeoxA/Tox where eo is the permittivity in a vacuum, Cox is the oxide capacitance, A is the cross-sectional area of oxide, and Tox is the oxide thickness,the dielectric constant of the oxide, eox, was calculated to be 10.5, a value which is in agreement with the tabulated value for MgO (9.8). The interface state density was calculated using the Terman method to be 4x1011eV-1cm-2 at 0.3 eV below the conduction band edge, as shown in Figure 4. The Terman method provides a rough evaluation of MgO/GaN interface trap density. It is based on the extraction of the experimental surface potential, ys, versus gate voltage, Vg. The measured CV curve stretched out away from the theoretical CV curve, which leads to the determination of interface trap density.The Terman method is generally found to under-estimate the trap density[10].

Figure 3. Typical C-V curve of a MgO/GaN diode under dark conditions at
room temperature.

The conductance technique was also used to characterize the MgO/GaN MOS
capacitors. The admittance of the MOS
capacitor measured is Gm + jwCm, where Gm
is measured equivalent parallel conductance, w is the angular frequency,
and Cm is the measured capacitance.
Converting the admittance to an impedance, subtracting the reactance of
the oxide capacitance, and converting back to an admittance yielded for the
real part
where Cox is measured in strong accumulation[12]. The estimated and theoretical fit of Gp/w were plotted versus log (w) for selected voltages of an 100 mm diameter MgO/GaN MOS diode at room temperature and are shown in Figure 4.

Figure 4. Conductance-frequency curves for selected gate bias voltage of a 100 mm diameter MgO/GaN MOS diode.
The width of the Gp/w versus log (w) curve depends only on the band-bending. The Gp/w versus frequency curve will be spread over a bias range determined by band-bending and Dit. For each Gp/w curve, the standard deviation of band-bending, ss, can be obtained. The Dit were calculated with
where fD is a function depending on s, A is the area of diode, and fp is the frequency at the maximum number of Gp/w. The values for fD from our samples ranged from 0.35 to 0.4. The estimated interface trap density is illustrated in Figure 5 along with the results obtained from Terman method. An interface state density of 4 ´ 1011/eV.cm2 at Ec-Et = 0.3 eV was obtained, which was slightly higher than that of Terman method. The Terman method provides a rough evaluation of MgO/GaN interface trap density and in general under-estimates of the trap density are expected[11]. Though the Terman method underestimates interface state density, its value is much more reasonable near the conduction band edge.

Figure 5. Interface trap
density distribution close to the GaN conduction band obtained on a MgO/GaN MOS
diode.
As described by Nicollian and Brews[11], the AC conductance technique is not able to measure interface traps with a time constant much longer than the period of the applied AC signal, because these traps cannot respond to the AC perturbation. High temperature AC conductance measurement were conducted at a temperature range of 25 to 300 °C to investigate the traps with longer time constants. An acceptor interface trap can become neutral or negative by accepting an electron and the distribution function, FSD, for the interface traps are similar to those for the bulk impurity level:

where Eit is the energy of the interface trap, k is the Boltzmann constant, T is the temperature, and g is the ground-state degeneracy, which is 4 for acceptor. The interface trap density linearly increased from room temperature to 200 °C and slightly level at 300 °C. The interface trap density was almost three times larger at 300 °C as compared to that at room temperature. This may have significant impact on the MOSFET performance at elevated temperatures.

Figure 6. The interface trap density
measured at different temperatures.
GaN MOS diodes were demonstrated using MBE grown MgO as the insulator. The breakdown fields of MgO diodes were 1.2 MV/cm. From the C-V measurement, showed good charge modulation from accumulation to depletion and deep depletion was observed for the room temperature measurements. Both Terman and AC conductance were used to estimate the interface state density and an interface state density of 4 ´ 1011/e-v.cm2 at Ec-Et = 0.3 eV was obtained. Elevated temperature conductance measurements were also performed. The interface trap density linearly increased from room temperature to 200 °C and started to saturate at 300 °C. The interface trap density increased to 6 ´ 1011/eV.cm2 at 300 °C.
Acknowledgments
This work was supported by ONR (contract No. N00014-98-1-0204) and
monitored by Dr. J. C. Zolper.