Characteristics of δ-doped InAlAs/InGaAs/InP high electron mobility transistors with a linearly
graded In
x Ga1−x As channel
This article has been downloaded from IOPscience. Please scroll down to see the full text article.
2006 Semicond. Sci. Technol. 21 619
(http://iopscience.iop.org/0268-1242/21/5/009)
Download details:
IP Address: 140.116.208.56
The article was downloaded on 07/10/2009 at 06:54
Please note that terms and conditions apply.
The Table of Contents and more related content is available
Semicond. Sci. Technol. 21 (2006) 619–625 doi:10.1088/0268-1242/21/5/009
Characteristics of
δ
-doped InAlAs/
InGaAs/InP high electron mobility
transistors with a linearly graded
In
x
Ga
1
−x
As channel
Jun-Chin Huang
1, Wei-Chou Hsu
1, Ching-Sung Lee
2,
Dong-Hai Huang
1and Ming-Feng Huang
11Institute of Microelectronics, Department of Electrical Engineering, National Cheng Kung
University, 1, University Road, Tainan, Taiwan, Republic of China
2Department of Electronic Engineering, Feng Chia University, 100 Wenhwa Road, Taichung,
Taiwan 40724, Republic of China E-mail:wchsu@eembox.ncku.edu.tw
Received 30 September 2005, in final form 23 December 2005 Published 21 March 2006
Online atstacks.iop.org/SST/21/619
Abstract
Various static and microwave performances on InAlAs/InGaAs/InP
HEMTs with a linearly-graded InxGa1−xAs channel (LGC-HEMT) have
been comprehensively investigated and compared to those having a
conventional lattice-matched In0.53Ga0.47As channel (LM-HEMT).
Improved carrier transport characteristics and confinement capability by employing the linearly-graded channel have contributed to superior extrinsic
transconductance (gm) of 346 mS mm−1, gate-voltage swing (GVS) of 0.5 V
(182 mA), unity-gain cut-off frequency (ft) of 41 GHz and maximum
oscillation frequency (fmax) of 63 GHz, with an improved frequency
operation plateau at 300 K for a gate dimension of 0.65× 200 µm2.
Furthermore, improved kink effects leading to a lower gate leakage
current of 0.7 µA mm−1, lower output conductance (gd) of 3.6 mS mm−1,
higher voltage gain (AV) of 93.1, higher off-state breakdown voltage of
16.3 V and superior output power characteristics have also been discussed.
1. Introduction
Over the past few years, InP-based HEMTs have demonstrated high-frequency and low-noise circuit applications due to their low effective electron mass, high low-field electron mobility, high electron saturation velocity and high sheet carrier densities in the InGaAs channel, as compared to those of GaAs-based HEMTs [1–3]. Nevertheless, low energy-gap InGaAs compounds are usually accompanied by low impact-ionization threshold fields and kink effects, thus considerably degrading the device performance, such as higher gate leakages, increased output conductance and decreased off-state breakdown voltages.
To resolve these problems, several approaches have been used, such as using an InP surface passivation layer over the
InAlAs gate recess regions to suppress the ionized-hole current injected to the gate terminal to improve breakdown voltages [4,5]. In addition, the InP layer is adopted as the etching-stop layer to precisely maintain the same Schottky layer thickness and to improve the uniformity of the device threshold and the extrinsic transconductance characteristics. Another way was to reduce the indium composition in the InGaAs channel to directly increase the impact-ionization threshold fields [6].
In this work, we present a δ-doped InAlAs/InGaAs/InP linearly-graded InxGa1−xAs channel high electron mobility transistor with an InP etching-stop layer. The linearly-graded channel structure can provide advantages of higher mobility characteristics and improved carrier confinement at the In0.56Ga0.44As-channel/spacer interface, and can
J-C Huang et al i-InP Etch-Stopper i-InxGa1-xAs Channel i-In0.52Al0.48As Buffer Capper n+-In 0.53Ga0.47As i-In0.52Al0.48As Schottky δ(n+) i-In0.52Al0.48As Spacer
S.I. InP Substrate Gate (Pt/Au) Source (AuGeNi/Au) Drain (AuGeNi/Au) LM-HEMT x = 0.53 LGC-HEMT x = 0.56 ↑ 0.5 InP In0. 52 Al 0. 48 As In0. 5 2 Al 0. 4 8 As Inx Ga 1-x As δ(n+) ∆Ev ∆Ec In0.5 2 Al 0.4 8 As (b) (a)
Figure 1. (a) Schematic cross section and (b) conduction band
diagrams for the studied LGC-HEMT and LM-HEMT, respectively.
simultaneously improve the impact-ionization threshold fields within the wider-gap In0.5Ga0.5As channel/buffer interface. The devised structure demonstrates low gate leakages, high breakdown voltages to achieve superior dc and RF performances. Comprehensive comparisons of the device characteristics to those having a lattice-matched In0.53Ga0.47As channel have also been discussed.
2. Material growth and device fabrication
The studied structures were grown by the low-pressure metal organic chemical vapour deposition (LP-MOCVD) system on the Fe-doped semi-insulating InP substrates. Figure 1(a) shows the schematic cross section of the epitaxial structure for the δ-doped In0.52Al0.48As/InxGa1−xAs/InP HEMT either
Table 1. Hall measurement results at 77 K and 300 K for the
LM-HEMT and LGC-HEMT, respectively.
LM-HEMT LGC-HEMT Temperature 300 K 77 K 300 K 77 K
µ(cm2V−1s−1) 8491 27 865 8845 31 523
ns(×1012cm−2) 3.5 3.1 3.77 3.35
µ–ns(×1016V−1s−1) 2.97 8.64 3.33 10.6
with a lattice-matched channel (LM-HEMT) or with a linearly-graded channel (LGC-HEMT). The same 50 nm thick In0.52Al0.48As buffer layers were grown on the S. I. InP substrate in both samples. Upon the buffer, a 12 nm thick undoped and compositionally-graded InxGa1−xAs channel layer, with x= 0.5–0.56 linearly increasing from the channel/buffer interface to the spacer/channel interface, has been grown for the LGC-HEMT, and a same 12 nm thick uniformly lattice-matched In0.53Ga0.47As channel layer has been deposited for the LM-HEMT. Besides, both channel structures provide the same averaged indium composition. A 5 nm thick undoped In0.52Al0.48As spacer, followed by the silicon planar doping layer (4 × 1012 cm−2), a 15 nm thick undoped In0.52Al0.48As Schottky layer, a 2.5 nm thick undoped InP layer and finally a 50 nm thick Si-doped (1 × 1019 cm−3) In
0.53Ga0.47As capper were sequentially grown on both samples. The undoped InP layer was inserted to serve as a gate-recess etching-stopper to improve the gate leakages, the output conductance (gd) and the voltage gain (Av). Figure 1(b) illustrates the corresponding band diagram of the studied devices at thermal equilibrium, where Ec is the conductance band edge, Evis the valence band edge and Efis the energy of the Fermi level.
Standard photolithography, lift-off and the rapid thermal annealing techniques were employed for both device fabrications. AuGe/Ni alloys were used for the source and drain ohmic contacts, onto which Au was evaporated to reduce the contact resistance. Gate recess was performed by employing the H3PO4/H2O2/H2O selective etching solution between the InGaAs capper and the InP layer. Pt/Au alloys were deposited on the undoped InP Schottky layer as the gate electrode. The gate length was 0.65 µm with the drain-to-source spacing of 4 µm. Mesa etching was further performed down to the buffer layer to reduce the substrate leakages.
3. Experimental results and discussions
Hall measurements have been conducted after removing the cap layers of the device structures under a magnetic field of 5000 G. Table 1 lists the Hall measurement results of the studied LM-HEMT and LGC-HEMT, including the two-dimensional electron gas (2DEG) concentration (ns), the electron mobility (µn) and the ns–µn product at 300 K (77 K), respectively. Owing to the narrower bandgap at a higher indium composition in the InGaAs channel resulting in the larger conduction band discontinuities (EC) of 0.54 eV at the spacer/channel interface in LGC-HEMT, which is higher than that of LM-HEMT, the linearly-graded channel structure provides superior transport properties and better carrier confinement capability than those of the LM-HEMT. 620
Characteristics of δ-doped InAlAs/InGaAs/InP high electron mobility transistors 0 1 2 Drain-Source Voltage (V) 0 100 200 300
Drain Current Density (mA/mm)
LM-HEMT LGC-HEMT
VGS = 0.5 V to -0.75 V
-0.25 V/step
3
Figure 2. The common-source I–V characteristic of the LM-HEMT
and LGC-HEMT at 300 K.
Figure 2 shows the common-source current–voltage characteristics of the LGC-HEMT and LM-HEMT at 300 K. It is evident that the LGC-HEMT has demonstrated a higher current density than the LM-HEMT, due to its improved transport property and carrier concentration. In addition, it is because the high resistivity and the wide-gap InAlAs buffer layer significantly suppress the buffer leakages that good pinch-off characteristics have been achieved in both devices. In the saturation region, the drain–source currents of the LGC-HEMT have demonstrated improved kink effects and lower output conductance than the LM-HEMT, since the LGC-HEMT has a higher threshold field to effectively suppress the impact-ionization phenomenon than the LM-HEMT, as the carriers were pushed towards the channel/buffer interface under the increased biases.
Figure 3 shows the extrinsic transconductance characteristics as a function of the saturated drain-current densities, and the inset indicates the dependences of both the extrinsic transconductance and the saturated drain-current density on the applied gate bias for both the LM-HEMT and LGC-HEMT at 300 K, respectively. The maximum extrinsic transconductance values (gm,max) at VDS= 2 V are 346 and 295 mS mm−1for the LGC-HEMT and LM-HEMT, respectively. Superior transport property at a higher In-composition channel regime near the gate electrode and the improved 2DEG confinement due to the increased channel/spacer discontinuities of the LGC-HEMT have contributed to the higher gm,max and higher current density performances than those of the LM-HEMT. Define the gate-voltage swing (GVS) as the width of the transconductance plateau of a 10% reduction from the value of gm,max. The LGC-HEMT has demonstrated an improved GVS of 0.5 V with the corresponding current regime of 191 mA (40 mA IDSS 231 mA) compared to 0.35 V and 127 mA (50 mA IDSS 171 mA) in the LM-HEMT, respectively, as shown in figure 3. The wider and higher extrinsic transconductance
0 100 200 300
Drain Current Density (mA/mm)
0 100 200 300 400 Extrinsic Transconductance (mS/mm) LGC-HEMT LM-HEMT VDS = 2 V -0.8 -0.4 0 0.4 0.8 Gate-Source Voltage(V) 0 100 200 300 400 Extrinsic Transconductance (mS/mm) 0 100 200 300 S a tu r a ti o n D r a in C u r r e n t Density (mA /mm) LM-HEMT LGC-HEMT
Figure 3. Extrinsic transconductance versus drain saturation density
for LGC-HEMT and LM-HEMT at 300 K, respectively. The inset shows the dependences of both the extrinsic transconductance and the drain saturation density on the applied gate bias at 300 K, respectively.
plateau of the LGC-HEMT was observed compared to the LM-HEMT. The improved GVS linearity is mainly attributed to the built-in electric fields, resulted from the tilted conduction band diagram within the linearly-graded InxGa1−xAs channel as shown in figure 1(b), retard the electrons being pushed from the higher In-composition channel regime towards the lower In-composition channel/ buffer interface at decreased gate biases. The resulted good device linearity in the LGC-HEMT can significantly reduce the inter-modulation of high-frequency signals, and further improve the distortion problem for high-power applications.
Figure4indicates the gate leakage current as a function of gate bias at different drain–source voltages. The gate leakage current consists of two major contributions: (1) electrons, tunnelling from the Schottky gate to the channel, result in a steady increase of IGwith the decreasing gate–drain biases (VGD), and (2) holes, generated by the impact ionization in the InGaAs channel, tunnel through the valence band barrier to the gate contact. The second contribution is approximately proportional to the drain current and can provide a clear indication of the kink effects. Since in the high drain-current region, a high electric field is established within the gate–drain regime at high VDS, electrons acquire enough energy to initiate the impact ionization to further generate electron–hole pairs (EHPs) to exaggerate the kink effects. As shown in figure4, the on-state peak gate leakage current of LM-HEMT is 0.7 (4.6) µA mm−1of the LGC-HEMT (LM-HEMT). The LGC-HEMT has demonstrated significantly improved kink effects than the LM-HEMT. Since the higher In-composition InGaAs compounds will have the higher impact-ionization coefficients, the LGC-HEMT was designed to have a lower indium composition at the channel/buffer interface to improve its high-field kink effects. The electrons, as depleted farther away from the gate at decreased gate biases, will move to
J-C Huang et al (a) VDS = 1 V to 3 V 0.5 V/step -0.8 -0.6 -0.4 -0.2 0 Gate-Source Voltage (V) -5 -4 -3 -2 -1 0 LM-HEMT VDS = 1 V to 3 V 0.5 V/step -0.8 -0.6 -0.4 -0.2 0 -0.8 -0.6 -0.4 -0.2 0 LGC-HEMT
Gate Current Density (uA/mm)
Gate Current Density (uA/mm)
Gate-Source Voltage (V) (b)
Figure 4. The on-state gate currents as a function of gate-source
voltage at various drain–source biases for (a) the LM-HEMT and (b) the LGC-HEMT, respectively.
the InGaAs channel region with lower In-composition and wider energy gap and, consequently, greatly improve the impact-ionization-related kink effects. Moreover, the InP layer, having also increased the valence band discontinuity by about 0.23 eV, can strongly reduce the impact-ionized holes tunnelling to the gate electrode. Both the LGC-HEMT and LM-HEMT have exhibited superior low peak on-state gate leakages as compared to the previous reports [7–10].
The characteristics of extrinsic transconductance (gm), output conductance (gd) and voltage gain (AV) versus the drain–source voltages are shown in figure5. It is noted that no peak values in the gdversus VDScharacteristics were observed, possibly due to the design of inserting an InP passivation layer over the InAlAs gate recess region to effectively suppress the ionized-hole current injected to the gate terminal, as discussed previously. Under the bias condition of VDS = 2 V and VGS= 0 V, with an identical density of IDS= 110 mA mm−1
gm Av gd 0 1 2 3 Drain-Source Voltage (V) 0 100 200 300 400 500 Extrinsic Transconductance gm (mS /mm)/ Output Conductance g d (mS/mm) 0 20 40 60 80 100 V o lta ge G a in A v LM-HEMT LGC-HEMT
Figure 5. Extrinsic transconductance, output conductance and
voltage gain as a function of drain–source voltage at VGS= 0 V for
the LGC-HEMT and LM-HEMT at 300 K, respectively.
for both devices, gdare 3.6 mS mm−1and 9.3 mS mm−1for the LGC-HEMT and LM-HEMT, respectively. As illustrated in figure 5, the LGC-HEMT presents superiorly low and flat output conductance at high VDSbiases. On the other hand, the gd value of the LM-HEMT slightly increases with the increased VDS, instead, owing to the different channel designs. Besides, lower output conductance in the saturation region can provide higher output impendence, beneficial to the voltage gain performance and relieving the loading effects. The calculated AVvalues, defined as gm/gd, are 93.1 and 30.8 for the LGC-HEMT and LM-HEMT, respectively. The LCG-HEMT has demonstrated about four times improvement than the LM-HEMT.
The two-terminal gate–drain breakdown characteristics at room temperature of the studied HEMTs are shown in figure 6. The two-terminal gate–drain breakdown voltages (BVGD), defined at IGD = −1 mA mm−1 are 16.6 V and 18.25 V for the LM-HEMT and LGC-HEMT, respectively. The device breakdown characteristics usually depend on the combinational effects of the gate tunnelling, the thermionic-field emission and the impact-ionization mechanisms in the channel [9]. Since the impact-ionization effect in the channel of the LGC-HEMT is improved upon that of the LM-HEMT, higher breakdown voltages of the LGC-MHEMT have been achieved over the LM-HEMT. In addition, the BVDG performance of this work is significantly superior to the 8.35 V of the composite channel HEMT [9], the 14.8 V of the lattice-matched channel HEMT [10], the 10.7 V of the pseudomorphic-channel HEMT [11], the 9.5 V of the step-graded channel HEMT [12], the 10.5 V of the triple-channel HEMT [15] and the 8.35 V of pseudomorphic-channel MHEMT [16] with the comparable gate lengths. The inset in figure6indicates the device forward turn-on voltage (Von) characteristics. The values of Von, defined at IGD = 1 mA mm−1, are 0.5 V and 0.8 V for the LM-HEMT and LGC-HEMT, respectively. Note that the turn-on voltage indicates the available forward gate bias before the gate leakage degradation occurs. A higher Vonvalue means a larger 622
Characteristics of δ-doped InAlAs/InGaAs/InP high electron mobility transistors -20 -15 -10 -5 0 5 Gate-Drain Voltage (V) -2 -1 0 1 2
Gate Current Density (mA/mm) LM-HEMT
LGC-HEMT 0 0.2 0.4 0.6 0.8 1 Gate-Drain Voltage (V) 0 0.4 0.8 1.2 1.6 2 G a te C u rrent D ensit y ( m A /m m )
Figure 6. Two-terminal gate-to-drain breakdown characteristics for
the LGC-MHEMT and LM-HEMT at 300 K. The inset shows the zoomed-in forward bias characteristics.
-3 -2 -1 0 Gate-Source Voltage (V) 0 2 4 6 8 10 12 14 16 18 -1 -0.8 -0.6 -0.4 -0.2 0
Gate Current Density (mA/mm)
LM-HEMT LGC-HEMT
ID = 1 mA/mm
BVoff
Drain-Source Voltage (V)
Figure 7. The drain–source off-state breakdown characteristics of
the LGC-MHEMT and LM-HEMT at 300 K.
current level in the channel may be induced at the extended forward gate bias, resulting in an enhanced device power handling capability and an increased operation range. The improvement of Vonof the LGC-HEMT on the LM-HEMT is believed to be attributed to the increased conduction band discontinuity at the spacer/channel interface to improve the forward gate leakages.
In addition, the maximum output power is determined by the maximum applied bias, i.e., the off-state voltage, and the available current swing that the device can sustain. Therefore, the off-state breakdown voltage is an important parameter for power application and can be determined by using the drain-current injection technique. Figure 7 indicates the
0 100 200 300
Drain Current Density (mA/mm) -20 0 20 40 60 80 Frequency (GHz) LM-HEMT LGC-HEMT fmax ft ft fmax
Figure 8. The ftand fmaxdependences on the drain current of the
LGC-HEMT and LM-HEMT at 300 K, respectively.
off-state drain–source breakdown characteristics as a function of the applied gate voltage with a fixed current injection of 1 mA mm−1into the drain terminals of the LM-HEMT and LGC-HEMT at 300 K. The off-state breakdown voltage (BVoff) is defined to be at the peak of the VDS curve and at the extrapolation point of IG = −1 mA mm−1. BVoff characteristics are determined to be −14.35 V for the LM-HEMT, and −16.3 V for the LGC-HEMT at 300 K. As shown in figures 6 and 7, the LGC-HEMT demonstrates superior Von, BVGD and BVoff performances to those of the LM-HEMT, due to the increased conductance band discontinuity and the suppressed impact-ionization effects by using the linearly-graded channel design. The LGC-HEMT is promising to provide favourable power application over the LM-HEMT.
The microwave on-wafer S-parameter measurements have also been conducted from 0.5 to 40 GHz in a common-source configuration, by using the HP-8510B network analyser at 300 K. The fT and fmax performances versus IDSat VDS= 2 V have been indicated in figure8. The maximum value of measured unity-gain cut-off frequency (ft) is 41 (32) GHz and the maximum oscillation frequency (fmax) is 63 (41) GHz at VDS= 2 V and VGS= 0 V for the LGC-HEMT (LM-HEMT), respectively. The higher fT and fmax characteristics of the LGC-HEMT have been achieved due to the intrinsic high-speed property of the narrower bandgap channel near the gate electrode and the improved output conductance. Furthermore, the LGC-HEMT has also shown an improved ft plateau with a wider drain-current regime, from 75 mA mm−1 to 225 mA mm−1, to maintain ft over 30 GHz, as compared to those of the LM-HEMT.
The extracted values and their respective physical meanings of the external parasitics in the small-signal equivalent circuit of the studied InP HEMTs, including Lg, Ld, Ls, Rg, Rd and Rs, have been defined in table 2. The external parasitic parameters were determined at zero drain-to-source bias by using the characterization method [17,18]. The ‘hot’-measured S-parameters (VDS>0 V) can then be further extracted from the external parasitic parameters to provide
J-C Huang et al
Table 2. Optimum extracted parasitics and small-signal parameters
for the LM-HEMT and LGC-MHEMT, respectively. Extracted parasitics and
small-signal parameters LGC-HEMT LM-HEMT Gate inductance, Lg(nH) 0.021 0.020 Drain inductance, Ld(nH) 0.1 0.095 Source inductance, Ls(nH) 0.016 0.01 Gate resistance, Rg() 22.1 22.3 Drain resistance, Rd() 13.1 13.0 Source resistance, Rs() 3.7 3.8 Charging resistance, Ri() 2.7 3.1 Output resistance, Rds() 874 415 Output conductance, gd(mS mm−1) 5.7 12.0 Intrinsic transconductance, 420 335 Gm(mS mm−1) Transconductance delay, T (ps) 1.4 1.6 Gate-drain capacitance, Cgd(pF) 0.0146 0.0162 Gate-source capacitance, Cgs(pF) 0.281 0.296 Drain-source capacitance, Cds(pF) 0.0825 0.0841
the equivalent Y-parameters of the intrinsic devices, including Cgd, Cgs, Cds, Ri, Rds, Gmand T [18], as also defined in table2. The extrinsic transconductance can be related to the intrinsic transconductance by gm= Gm× exp(–jωT), where ω is the frequency in radians. Table 2 lists the optimum extracted parameters of the small-signal equivalent circuit at VDS= 2 V and VGS= 0 V, with an identical current density of IDS= 110 mA mm−1, for the LM-HEMT and LGC-HEMT, respectively. As shown in table 2, high intrinsic transconductance, Gm= 420 (335) mS mm−1, and low output conductance, gd= 5.5 (12.1) mS mm−1, of the LGC-HEMT (LM-HEMT) have contributed to a superior intrinsic voltage gain (Gm/gd) of 76.36 (27.68). Consistent improvements on the kink effects and the voltage gain of the LGC-HEMT over the LM-HEMT have also been verified from the high-frequency parametric extraction, as expected from the static characterizations discussed in the previous section. The intrinsic ft can be estimated to be 45.2 (34.1) GHz for the LGC-HEMT (LM-HEMT) through the approximations, ft ≈ 2π(Cgm
gd+Cgs), where Cgsis the capacitance between gate
and source, Cgdis the capacitance between gate and drain, Ri is the channel resistance and gdis the output conductance.
Microwave power characteristics were also investigated by using a load-pull ATN system, which provides a conjugate simultaneously matched input and load impedances for achieving an optimum power performance. The microwave power performances are measured at 5.8 GHz, with VDS= 2 V and VGS = −0.25 V at 300 K. Figure 9 shows the saturated output power (Pout), small-signal power gain (Gs) and the power-added efficiency (PAE) versus the input power for both the LGC-HEMT and the LM-HEMT, respectively. The devices were operated under the class-AB condition, and compromised by the power-added efficiency (PAE) and the output power. The measured saturated output power, PAE and the associated power gain are 11.04 (12.44) dBm, 36.9 (41.14)% and 18.51 (20.07) dB at VGS= −0.25 V for the LM-HEMT (LGC-HEMT), respectively. Together with the superior breakdown characteristics and current driving capability, the improved high-power performances of the
-20 -10 0 -5 0 5 10 15 20 25 0 10 20 30 40 50
Power Added Efficiency PAE (%)
LM-HEMT LGC-HEMT Output Power Power Gain PAE Input Power (dBm)
Power Gain(dB)/Output Power (dBm)
Figure 9. Room-temperature output power, power gain and
power-added efficiency characteristics versus input power for the LGC-HEMT and LM-HEMT at 5.8 GHz, with VDS= 2 V and
VGS= −0.25 V, respectively.
LGC-HEMT indicate its promising power applications over the LM-HEMT.
The noise characteristics have also been measured on-wafer by using an HP 8970B noise meter in conjunction with the cascade probes over the frequency range from 1 GHz to 10 GHz. The minimum noise figures (NFmin) and the associated gain for the LGC-HEMT (LM-HEMT) were determined to be 0.65 (0.83) dB and 15.65 (13.32) dB, respectively, at 5.8 GHz with VDS = 2 V and VGS = −0.25 V. Since the impact ionization in the channel would possibly cause additional shot noise contributions, better noise characteristics have been observed in the LGC-HEMT than the LM-HEMT mainly due to its improved impact-ionization effects and the device gain by using the linearly-graded channel design.
4. Conclusion
In conclusion, a δ-doped In0.52Al0.48As/InxGa1−xAs LGC-HEMT has been successfully grown by the low-pressure metal organic chemical vapour deposition (LP-MOCVD) system. Influences on the various dc and high-frequency device characteristics of the LGC-HEMT and LM-HEMT have been comprehensively discussed. The distinguished linearly-graded InxGa1−xAs channel structure improves the impact-ionization effects in the channel and diminishes the kink effects, resulting in the lower gate leakage currents, the higher current drive, the lower output conductance, the higher off-state breakdown voltages and the higher voltage gains. Furthermore, the improved carrier transport characteristics, the enhanced confinement capability and the devised built-in retarding electric field by employing the linearly-graded channel design have contributed to the wider gate-voltage swing, the improved output power and noise performances, the higher unity-gain cut-off frequency and the higher maximum oscillation frequency (fmax n) with 624
Characteristics of δ-doped InAlAs/InGaAs/InP high electron mobility transistors wider frequency operation plateaus, as compared to those
in the LM-HEMT. The results demonstrate that the LGC-HEMT is promising for high-linearity and high-power circuit applications.
Acknowledgment
This work was also supported by the National Science Council of the Republic of China under the contract numbers of NSC 93-2215-E-006-007, NSC 2215-E-035-012 and NSC 94-2815-C-020-E.
References
[1] Yamashita Y, Endoh A, Shinohara K, Hikosaka K, Matsui T, Hiyamizu S and Mimura T 2002 IEEE Electron Device
Lett.23 573
[2] Medjdoub F, Zaknoune M, Wallart X, Gaquiere C and Theron D 2004 IEEE Electron Device Lett. 41 241 [3] Shinohara K, Yamashita Y, Endoh A, Watanabe I, Hikosaka K,
Matsui T and Hiyamizu S 2004 IEEE Electron Device Lett.
25 241
[4] Suemitsu T, Yokoyama H, Ishii T, Enoki T, Meneghesso G and Zanoni E 2002 IEEE Trans. Electron Devices49 1694
[5] Meneghesso G, Buttari D, Perin E, Canali C and Zanoni E 1998 IEDM Tech. Dig. p 227
[6] Boudrissa M, Delos E, Gaquiere C, Rousseau M, Cordier Y and Theron D 2001 IEEE Trans. Electron Devices
48 1037
[7] Heedt C, Buchali F, Prost W, Brockerhoff W, Fritzsche D, Nickel H, Losch R and Tegude F J 1994 IEEE Trans.
Electron Devices41 1685
[8] Somerville M H, Alamo J A and Hoke W 1996 IEEE Electron
Device Lett.17 473
[9] Menehesso G, Neviani A, Oesterholt R, Matloubian M, Lur T, Brown J J, Canali C and Zanoni E 1999 IEEE Trans.
Electron Devices46 2
[10] Kim D H, Yeon S J, Lee J H and Seo K S 2003 IEDM Tech.
Dig. p 30.5.1
[11] Chen Y W, Hsu W C, Hsu R T, Wu Y H and Chen Y J 2004
Solid-State Electron.48 119
[12] Chen Y J, Hsu W C, Chen Y W, Lin Y S, Hsu R T and Wu Y H 2005 Solid-State Electron.49 163
[13] Ao J P, Zeng Q M, Zhao Y L, Li X J, Liu W J, Liu S Y and Liang C G 2000 IEEE Electron Device Lett.21 200
[14] Mahajan A, Arafa M, Fay P, Caneau C and Adesida I 1997
IEEE Electron Device Lett.18 284
[15] Maher H, D´ecobert J, Falcou A, Post G and Scavennec A 1999
Int. Proc. Conf. Indium Phosphide and Related Materials
p 315
[16] Hsu W C, Chen Y J, Lee C S, Wang T B, Lin Y S and Wu C L 2005 IEEE Electron Device Lett.26 59
[17] Dambrine G, Cappy A, Heliodore F and Playerz E 1988 IEEE
Trans. Microw. Theory Tech.36 1151
[18] Berroth M and Bosch R 1991 IEEE Trans. Microw. Theory
