Drain Gate
K. Mishra, Solid State Electron., 41, 1569 (1997)
[15] R. Gaska, J. Yang, A. Osinsky, M. F. Khan and M. S. Shur, IEDM Tech. Dig., 565 (1997).
i-AlGaN 18 nm (Al=0.17)
i-GaN 3 µm Buffer layer
Sapphire
Fig. 6-1 The schematic of undoped structure (sample No. 1), where Al composition is 0.17 and the top AlGaN layer thickness is 18 nm.
i-AlGaN 50 nm (Al=0.17)
i-GaN 3 µm Buffer layer
Sapphire
Fig. 6-2 The schematic of undoped structure (sample No. 2), where Al composition is 0.17 and the top AlGaN layer thickness is 50 nm.
i-AlGaN 28 nm (Al=0.3)
i-GaN 3 µm
Buffer layer
Sapphire
Fig. 6-3 The schematic of undoped structure (sample No. 3), where Al composition is 0.3 and the top AlGaN layer thickness is 28 nm.
i-AlGaN 5 nm n-AlGaN: 5E18 20 nm
i-AlGaN 3 nm i-GaN 3 µm
Buffer layer
Sapphire
Fig. 6-4 The schematic of modulation-doped structure (sample No. 4), where Al composition is 0.3 and the AlGaN layer consists of a 3 nm undoped AlGaN spacer, a 20 nm AlGaN with Si doping concentration of 5×1018 cm-3 and a 5 nm undoped AlGaN cap layer.
Fig. 6-5 The surface morphology of undoped structure (sample No. 1), where Al composition is 0.17 and the top AlGaN layer thickness is 18 nm.
Fig. 6-6 The surface morphology of undoped structure (sample No. 2), where Al composition is 0.17 and the top AlGaN layer thickness is 50 nm.
Fig. 6-7 The surface morphology of undoped structure (sample No. 3), where Al composition is 0.3 and the top AlGaN layer thickness is 28 nm.
Fig. 6-8 The surface morphology of modulation-doped structure (sample No. 4), where Al composition is 0.3 and the total AlGaN layer thickness is 28 nm.
100 200 300 400 500
Fig. 6-9 A comparison on the electron concentration, where blank square ( )
represents the sample with Al=0.17 (sample No. 1) and circle (•) represent the sample with Al=0.3 (sample No. 3).
100 200 300 400 500
0
Fig. 6-10 A comparison on electron mobility, where blank square ( ) represents the sample with Al=0.17 (sample No. 1) and circle (•) represents the sample with Al=0.3 (sample No. 3)
100 200 300 400 500
Fig. 6-11 A comparison on electron concentration, where blank square ( ) represents the sample with AlGaN thickness of 18 nm (sample No. 1) and circle (•) represents the sample with AlGaN thickness of 50 nm (sample No. 2)
100 200 300 400 500
1000
Fig. 6-12 A comparison on electron mobility, where blank square ( ) represents the sample with AlGaN thickness of 18 nm (sample No. 1) and circle (•) represents the sample with AlGaN thickness of 50 nm (sample No. 2).
100 200 300 400 500
Fig. 6-13 A comparison on electron concentration, where blank circle ( ) represents the undoped structure (sample No. 3) and circle (•) represents modulation-doped structure (sample No. 4).
100 200 300 400 500
500
Fig. 6-14 A comparison on electron mobility, where blank circle ( ) represents the undoped structure (sample No. 3) and circle (•) represents modulation-doped structure (sample No. 4)
10 15 20 25 30 35 40 45 50
Fig. 6-15 The carrier profile of the undoped structure (sample No. 3) by C-V measurement. The dash line represents the location of AlGaN/GaN interface.
10 15 20 25 30 35 40 45 50
Fig. 6-16 The carrier profile of the modulation-doped structure (sample No. 4) by C-V measurement. The dash line represents the locations of AlGaN/GaN interface and undoped AlGaN spacer/Si-doped AlGaN interface, respectively.
Fig. 6-17 The SEM images of an undoped Al0.3Ga0.7N/GaN HFET with a gate length of 0.15 µm (top) and the enlarged picture of a 0.15 µm long narrow T-gate.
0 2 4 6 8 10
Fig. 6-18 Current-voltage characteristics of the undoped Al0.3Ga0.7N/GaN HFET. The gate length is 0.15 µm and the device width is 75 µm.
Fig. 6-19 DC transfer characteristics of the undoped Al0.3Ga0.7N/GaN HFET measured at Vds= 5 V. Its gate length is 0.15 µm and device width is 75 µm.
-120 -100 -80 -60 -40 -20 0
Fig. 6-20 Breakdown characteristics of the undoped Al0.3Ga0.7N/GaN HFET. Its gate length is 0.15 µm and device width is 75 µm.
Fig. 6-21 Forward Schottky I-V characteristics of the undoped Al0.3Ga0.7N/GaN HFET.
Its gate length is 0.15 µm and device width is 75 µm.
1 10 0
5 10 15 20 25 30
100
fmax
fT fT, fmax (dB)
frequency (GHz)
fT
fmax
undoped
Fig. 6-22 Small-signal characteristics of the undoped Al0.3Ga0.7N/GaN HFET. Its gate length is 0.15 µm and device width is 75 µm.
0 2 4 6 8 10
Fig. 6-23 Current-voltage characteristics of the modulation-doped Al0.3Ga0.7N/GaN HFET. The gate length is 0.15 µm and the device width is 75 µm.
-12 -10 -8 -6 -4 -2 0 2
Fig. 6-24 DC transfer characteristics of the modulation-doped Al0.3Ga0.7N/GaN HFET.
The gate length is 0.15 µm and the device width is 75 µm.
-120 -100 -80 -60 -40 -20 0
Fig. 6-25 Breakdown characteristics of the modulation-doped Al0.3Ga0.7N/GaN HFET.
Its gate length is 0.15 µm and device width is 75 µm.
Fig. 6-26 The forward Schottky I-V characteristics of the modulation-doped Al0.3Ga0.7N/GaN HFET. Its gate length is 0.15 µm and device width is 75 µm.
1 10
Fig. 6-27 Small-signal characteristics of the modulation-doped Al0.3Ga0.7N/GaN HFET. Its gate length is 0.15 µm and device width is 75 µm.
0.1 1
Fig. 6-28 The gate length dependence of the cut-off frequency for the undoped and modulation-doped Al Ga N/GaN HFETs.
Chapter 7
Thermal effect of AlGaN/GaN heterostructure Field Effect Transistors
7-1 Introduction
GaN-based heterostructure field-effect transistors (HFETs) are promising candidates for high-temperature and high-power applications at high frequencies due to their superior material properties, such as high breakdown field, high saturation velocity and excellent thermal stability [1-3]. For such operations, the stability of devices over temperature is extremely important. In addition to the commonly known problem in the thermal conductivity of substrates [4-5], the device structure plays a crucial role in realizing GaN-based HFETs for high-temperature applications. So far, several device structures, such as the undoped structure [6], the modulation-doped structure [7] and the channel-doped structure [8-9], have been used to realize high-performance GaN-based HFETs. Obviously devices with different structures may exhibit different electrical behaviors at high temperatures due to their different electron transport properties. In this article, we report a comparison on the temperature dependence of the electron transport properties and device characteristics of the undoped and modulation-doped AlGaN/GaN HFETs. The results obtained
indicate that the device structure has a significant influence on the electron transport properties of devices and the device performance at high temperatures. By the analysis of the activation energy, we identify that the increase in electron concentration in the modulation-doped structure at high temperatures is dominated by the thermal activation of Si donors. The modulation-doped devices, with a higher electron concentration, comparable mobility and lower parasitic source resistance at high temperatures, exhibited better dc and RF performance than the undoped devices over temperatures.
7-2 Experiments
Two structures, an undoped structure and a modulation-doped structure, were
grown by metalorganic chemical vapor deposition (MOCVD) on c-plane sapphire substrates. The undoped structure consists of a 3 µm undoped GaN buffer layer and a 28 nm undoped AlGaN cap layer. The modulation-doped structure consists of a 3 µm
undoped GaN buffer layer, a 3 nm undoped AlGaN spacer, a 20 nm Si-doped AlGaN layer with a doping concentration of 5×1018 cm-3 and a 5 nm undoped AlGaN cap layer. The Al composition is 0.3 for all samples. Contact metal, Ti/Al/Ti/Au (200/1500/450/550 Å), was deposited and annealed at 750°C for 30 s in N2 gas ambient to form ohmic contact pads. Hall effect measurements with Van der Pauw
geometry were performed to characterize the electrical properties of two-dimensional electron gases (2DEG) in the temperature range from 100K to 500K.
Capacitance-voltage (C-V) profiling revealed that most carriers were located at the AlGaN/GaN interface.
Device fabrication process began with mesa isolation, followed by Ohmic contact and narrow T-gate fabrication. Mesa patterns for device active regions were defined by photolithography and dry etch with Cl2/Ar plasma using an inductive coupled plasma (ICP) system. Prior to contact metal deposition, the samples were treated by Ar plasma. Contact metal, Ti/Al/Ti/Au (200/1500/450/550 Å), was then deposited and lifted-off to form the contact pads. The samples were annealed at 750°C for 30 s in N2 gas ambient. The contact resistances of 0.59 ohm-mm for the undoped
devices and of 0.38 ohm-mm for modulation-doped devices by the transmission line measurement (TLM) were obtained. The source-drain spacing is 2 µm for all samples.
In the narrow T-gate fabrication, the resist stack consisted of a low-sensitivity 495k PMMA layer of 120 nm thickness in the bottom and a high-sensitivity copolymer EL 12.5 PMMA-MAA (680 nm) on the top. The e-beam acceleration voltage used here was 15 kV. Metal deposition of Ni (20 nm)/Au (300 nm) was performed using e-beam evaporation. Finally, the sample was soaked in acetone for metal lift-off.
Device dc and small-signal s-parameter measurements were performed using
HP4142 dc analyzer and Anritsu37397 vector network analyzer (VNA) at temperatures ranging from room temperature to 200ºC to observe the device high temperature performance. Before measurement, the actual temperature of the surface of the hot chuck was checked carefully by a thermocouple. During high temperature measurement, a 15 minutes delay for probe stability was performed for a reliable high temperature s-parameter measurement.
7-3 Results and discussion
7-3-1 Electron transport properties
Figure 7-1 shows the temperature dependence of the electron concentration of the two structures. For the undoped structure, the sheet charge concentration of 2DEG is around 1×1013 cm-2 and is nearly constant over the measured temperature range.
This 2DEG is due to the strong spontaneous and piezoelectric polarization effect [10].
For the modulation-doped structure, the electron concentration is temperature dependent. The sheet charge concentration increases from 1.15×1013 cm-2 at 100K to 1.33×1013 cm-2 at 500K.
At very low temperatures, because of the freeze-out of Si donors in the AlGaN layer, the channel conduction of the modulation-doped structure markedly decreases.
As the temperature increases, the thermal activation process starts to release gradually
the frozen electrons and thus the measured sheet carrier density increases.
The activation energy of Si donors in the AlGaN layer was calculated using the temperature-dependent Hall data. The semi-log plot of the carrier density versus 1/T is shown in Fig. 7-2. The fitted line based on the charge neutrality condition [11],
)
is used to calculate the activation energy. In this equation, n is the increased electron concentration due to the thermal activation process; ND and NA are the donor and acceptor concentrations, respectively; Nc is the effective density of state in the conduction band; gd is the donor spin-degeneracy factor; and Ed is the activation energy of Si donors in Al0.3Ga0.7N. In this calculation, n is obtained using
AlGaN
where Ns(T) is the sheet charge concentration measured at different temperatures, the numerator, Ns(T)-Ns(100K), represents the increased sheet charge concentration due to the thermal activation process and dAlGaN is the effective Si-doped AlGaN layer thickness, a fitting parameter. Based on the fitted result, dAlGaN is estimated to be around 13 nm. This value is reasonable since part of the AlGaN layer is depleted due to the surface potential. ND is 5×1018 cm-3; NA is assumed to be much smaller than ND for the uncompensated condition; Nc is also temperature dependent and is proportional to T3/2; the donor spin-degeneracy factor, gd, is 2; and Ed is the fitting
parameter. Also, a single-donor model was used for this calculation. As can be seen in Fig. 7-2, a good fitting can be obtained with the activation energy of 83.2 meV. This result is slightly lower than the previously reported Si donor level in Al0.3Ga0.7N [12-13]. Thus, it is apparent that the thermal activation process in the Si-doped AlGaN layer plays an important role in the temperature dependence of the device performance for the modulation-doped structure.
Figure 7-3 shows the temperature dependence of the electron mobility of the two structures. As a whole, the electron mobilities of the two structures decrease with temperature as a result of the increased phonon scattering. The undoped structure has a low-temperature electron mobility of 3030 cm2/Vs at 100K and room-temperature electron mobility of 1100 cm2/Vs. The modulation-doped structure has an electron mobility of 2610 cm2/Vs at 100K and a room-temperature mobility of 953 cm2/Vs.
The lower mobility for the modulation-doped structure is due to the additional ionized impurity scattering associated with Si donors and possible parallel conduction in the doped AlGaN layer. As the temperature increased above 450K, the two structures showed comparable electron mobilities. At 500K, the mobilities are 537 cm2/Vs and 529 cm2/Vs for the undoped and modulation-doped structures, respectively. This is due to the enhanced phonon scattering that dominates over impurity scattering at high temperatures. Similar results were also obtained in a previous study [14].
7-3-2 Device high temperature performances
In this study, the actual temperature of the surface of the hot chuck is defined as the test temperature. Figure 7-4 shows the current-voltage characteristics of the
undoped AlGaN/GaN HFETs at room temperature and 200°C. The gate length and width are 0.15 µm and 75 µm, respectively. At room temperature, the undoped device
showed good dc performance. The maximum drain current is 700 mA/mm at drain bias Vds= 7 V and gate bias Vgs= 2 V. The reduction in the drain current at higher drain bias was due to the self-heating effect as a result of the poor thermal conductivity of the sapphire substrate. Good channel pinch-off and self-heating were also clearly observed at 200°C. The maximum drain current at 200°C was reduced to 567 mA/mm at Vds= 5.5 V and Vgs= 2 V. The drain current reduction at high temperature is due to the 2DEG mobility degradation.
Figure 7-5 shows the dc transfer characteristics of the undoped AlGaN/GaN HFETs at drain bias Vds= 5 V measured at different temperatures. In particular, the extrinsic transconductance showed small variation over a wide gate bias range, which indicated a good linearity of the undoped device. The maximum extrinsic transconductance at room temperature and 200°C is 113 mS/mm at Vgs= -3.8 V and 86 mS/mm at Vgs= -3 V, respectively. The lower extrinsic transconductance obtained
here is due to the high source resistance of the undoped device. The source resistance and the intrinsic transconductance at room temperature are 3.4 ohm-mm and 183 mS/mm, respectively. The threshold voltage is close to -7 V and is constant over the measured temperature range. This indicated the good thermal stability of the Schottky metal between the underlying undoped AlGaN in the undoped devices. Figure 7-6 shows the forward Schottky gate I-V characteristics of undoped Al0.3Ga0.7N/GaN HFETs at room temperature and 200ºC. At 200ºC, this device showed a slightly higher gate current than that at room temperature. The calculated ideality factor is around 2, which is slightly large. Thus the Schottky barrier height could not be extracted correctly. The gate resistance, which is extracted from the linear Schottky I-V, is very large, ~40000 ohm.
Figure 7-7 shows the current-voltage characteristics of the modulation-doped AlGaN/GaN HFETs at room temperature and 200°C. Device dimension is also 0.15×75 µm2. At room temperature, a very high maximum drain current of 1040 mA/mm is obtained at drain bias Vds= 7 V and gate bias Vgs= 2 V. The self-heating effect was also clearly observed at larger drain bias. The maximum drain current at 200°C is reduced to 678 mA/mm at Vds= 7 V and Vgs= 2 V.
Figure 7-8 shows the dc transfer characteristics of the modulation-doped AlGaN/GaN HFET at drain bias Vds= 5 V under different temperatures. The measured
extrinsic transconductance at different temperartures showed larger changes over the gate bias than the undoped devices. The maximum extrinsic transconductance at room temperature and 200°C is 198 mS/mm at Vgs= -6.3 V and 125 mS/mm at Vgs= -6.7 V, respectively. The source resistance and the intrinsic transconductance at room temperature are 2.67 ohm-mm and 420 mS/mm, respectively. The threshold voltage is close to -9 V and no obvious threshold voltage shift was observed over the whole temperature range. This indicated the good thermal stability of the Schottky metal between the underlying undoped AlGaN in the modulation-doped devices. Figure 7-9 shows the forward Schottky gate I-V characteristics of modulation-doped Al0.3Ga0.7N/GaN HFETs at room temperature and 200ºC. At 200ºC, this device also showed a slightly higher gate current than that at room temperature. The excellent thermal stability of the metal/nitride Schottky contacts over wide temperature was also observed in previous reports [15]. The gate resistance of modulation-doped device is around 25 ohm at room temperture, which is much smaller than the undoped device.
Figure 7-10 shows the comparison of the temperature dependence of the maximum drain current at a gate bias Vgs= 2 V. The undoped device showed the maximum drain current of 700 mA/mm at room temperature. At 200°C it reduced to 567 mA/mm. The modulation-doped device exhibited a larger maximum drain current.
At room temperature, it was of 1040 mA/mm and at 200°C it became 678 mA/mm.
The larger change in the maximum drain current for the modulation-doped device can
be attributed to the temperature dependent sheet carrier density. The undoped device showed a drain current changing rate of –5.8 mA/mm per 10°C. The
modulation-doped device, however, showed a drain current changing rate of –20.9 mA/mm per 10°C.
Figure 7-11 shows the comparison of the maximum extrinsic transconductances measured at a drain bias Vds= 5 V for both devices over temperature. In general, the modulation-doped devices had a higher transconductance than the undoped devices.
But the undoped device had a smaller change in transconductance over temperature.
For the modulation doped structure, the transconductance changed from 198 mS/mm at room temperature to 125 mS/mm at 200°C. For the undoped device, the transconductances ranged from 113 mS/mm to 86 mS/mm over the temperatures. The lower transconductance is due to the large source resistance, which was 3.4 ohm-mm, of the undoped channel in the undoped device. For the modulation-doped device it
was 2.67 ohm-mm. The change rate of the transconductance for the modulation-doped device was –4.4 mS/mm per 10°C.
Figure 7-12 shows the comparison of the temperature dependence of the current gain cut-off frequency (fT) for both devices. The undoped device was operated at Vds=
6 V and Vgs= -3.5 V. The modulation-doped device was operated at Vds= 6 V and
Vgs= -6 V. For the undoped device, the cut-off frequency was 32 GHz at room temperature, but degraded to 22 GHz at 200°C. The modulation-doped device had a
room temperature fT of 75 GHz and did not show obvious degradation until 100°C.
Above 100°C, the cut-off frequency became lower and dropped to 52 GHz at 200°C.
As a whole, both devices did not show obvious degradation for temperatures below 100°C. A similar result was also observed in the doped channel AlGaN/GaN HFETs
[8]. The main reason is the weak temperature dependence of electron transport property [16-17]. The lower cut-off frequency for the undoped device is mainly attributed to the larger parasitic source and drain resistances.
Based on the Hall measurement results, the undoped device had a constant two-dimensional electron gas (2DEG) concentration in the channel over a wide temperature range. The modulation-doped device had a higher electron concentration but it increased with temperature, due to the thermal activation of Si donors in the AlGaN layer. Although lower electron mobility was observed in the modulation doped devices due to the additional ionized impurity scattering associated with the Si donors, the electron mobilities for both devices are similar at high temperatures where phonon scattering is the dominant scattering process. Because of the additional doping, the modulation-doped devices has lower parasitic source and drain resistances than the
undoped devices over temperatures. Putting all these factors together, we may conclude that the modulation-doped devices are superior to the undoped devices over the temperatures studied. The stability (the temperature dependence of device performance), however, is not as good as the undoped devices.
7-4 Conclusion
We investigated the thermal effect of the AlGaN/GaN HFETs. We compared the temperature dependence of the electron transport properties and device characteristics of the undoped and modulation-doped AlGaN/GaN HFETs. The results obtained indicate that the device structure has a significant influence on the electron transport properties of devices and device performance over temperature. The undoped structure has a constant 2DEG concentration over a wide temperature range, while the modulation-doped structure has a temperature-dependent electron concentration. The increase in electron concentration in the modulation-doped structure at high temperatures was caused by the thermal activation of Si donors in the AlGaN layer.
Although the undoped structure has a higher mobility at low and moderate
Although the undoped structure has a higher mobility at low and moderate