In this letter, we design and demonstrate an improved metalorganic chemical vapor deposition (MOCVD) grown reverse Al-composition graded contact layer, whereby the Al-composition of AlxGa1−xN in the contact layer is graded from the higher Al-composition as in the channel to lower Al-composition, to achieve a low resistance contact to MOCVD grown ultrawide bandgap Al0.70Ga0.30N channel metal-semiconductor field-effect transistors. Increasing the thickness of the reverse graded layer was found to improve contact layer resistance significantly, leading to a contact resistivity of 3.3 × 10−5 Ω cm2. Devices with a gate length, LG, of 0.6 μm and a source-drain spacing, LSD, of 1.5 μm displayed a maximum current density, IDS,MAX, of 635 mA/mm with an applied gate voltage, VGS, of +2 V. Breakdown measurements on transistors with a gate to drain spacing, LGD, of 770 nm had breakdown voltage greater than 220 V, corresponding to a minimum breakdown field of 2.86 MV/cm—almost 3× higher than that exhibited by lateral GaN channel devices with similar dimensions. This work provides a framework for the design of low resistance contacts to MOCVD grown high Al-content AlxGa1−xN channel transistors.

High Al-composition AlxGa1xN with x 0.7 and bandgap EG ≥ 5.1 eV are estimated to have a high critical breakdown field, FBR, and high saturated velocity, vsat, which are attractive material properties for high voltage and high frequency power applications. This set of materials has FBR exceeding 11 MV/cm and correspondingly a high lateral figure of merit (LFOM),

LFOM=nsμFBR2,

where ns is the sheet charge density in the channel and μ is the carrier mobility.1 Since the saturated velocity vsat in AlxGa1xN is expected to be similar to GaN, the Johnson figure of merit (JFOM),

JFOM=vsatFBR2π,

for these materials is expected to be much higher than that for GaN.2–4 Devices fabricated with high Al-composition AlxGa1xN will thus be more suitable for scaling due to the high FBR. Compared to GaN, higher breakdown voltages are expected for devices with similar dimensions while maintaining comparable saturated current density, IDS,MAX, due to similar vsat.

However, establishing a low-resistance Ohmic contact to AlxGa1xN with x 0.7 is a challenge due to the low electron affinity in these materials. Although metal-semiconductor Ohmic contacts have been demonstrated for x >0.5, the lowest contact resistance achieved to date is 24.6 Ω mm.5–10 On the other hand, reverse Al-composition graded contact layers have been able to reduce contact resistance down to 0.3 Ω·mm for molecular-beam epitaxy (MBE) grown Al0.75Ga0.25N channel transistors and can be readily employed to form low resistance contacts, to any AlxGa1xN channel devices.11 In these structures, the composition of AlxGa1xN in the contact region is gradually varied from the channel composition to a lower composition, while the material is doped with [Si+] to ensure it remains n-type. The metal-semiconductor contact resistance is low in these cases since the contact is being made to GaN. Since the material is gradually graded, there are no abrupt semiconductor band offsets to impede electron transport. These molecular beam epitaxy (MBE) grown devices, however, have significantly lower mobility than theoretically predicted.12 The origin of the low mobility is still being investigated.13 In contrast, metalorganic chemical vapor deposition (MOCVD) grown high Al-composition AlxGa1xN channels have been shown to have much higher mobility due to superior compositional uniformity, with electron mobility exceeding 150 cm2/V s.5 Therefore, in this work, we focus on achieving low-resistance MOCVD-grown graded contact layers to MOCVD-grown high Al-composition AlxGa1xN channels.

We now discuss some considerations necessary for the design of the reverse-graded contact layer. The (Al)GaN material system displays a strong polarization effect due to the large piezoelectric polarization originating from the large ionicity of the (Al)Ga-N bonds and the strong spontaneous polarization resulting from the uniaxial nature of the crystal coupled with the nonideal c/a ratio of the wurtzite structure.14,15 Compositional grading of such polar materials produces bound polarization “bulk” charge density given by Dπ=×P, where P is the sum of spontaneous and piezoelectric polarization in AlxGa1xN alloys.15 When the Al-composition in AlxGa1xN is graded from a higher to lower Al-composition, a negative polarization charge is formed, causing a positive curvature in the energy band profile.11 To compensate for this negative polarization charge and obtain a flat conduction band profile, it is necessary to introduce heavy n-type doping via [Si+] incorporation in the lattice to produce a low resistance effective n-type region.11 The resistivity of the compositionally graded contact layer, ρgraded, can be described by

ρgraded=0tgradeddzqμNDzzPxz,
(1)

where q, μ, ND(z), and P[x(z)] are the elementary electronic charge, carrier mobility, impurity doping density in the film, and the polarization in the graded layer, respectively.

Doping of high Al-composition AlxGa1−xN films has been studied for both MBE and MOCVD growth techniques. Although reports of conductive high Al-composition AlxGa1−xN films exist for both methods, a discrepancy in the maximum achievable doping density has been observed between these two techniques.16–21 While studies with MBE grown high Al-content AlxGa1−xN films have consistently achieved a high doping density above mid-1019 cm−3 for x < 0.8, such reports have been inconsistent for MOCVD grown samples.16–21 The likely reason for this discrepancy is that for MOCVD grown AlxGa1−xN films, dopant incorporation has been found to be a strong function of the Al-composition of the film and high dopant incorporation becomes more challenging for higher Al-content AlxGa1−xN—which compounds the difficulty of growing reverse-graded contact layers for high Al-composition AlxGa1−xN.19,20 This is different from MBE-grown AlxGa1−xN films where dopant incorporation was found to be independent of the Al-composition of the film.16,17 Thus, while MBE can achieve a degenerately doped Al-composition graded contact layer using the same growth condition, a continuous shift in the optimum dopant incorporation condition makes growing such uniformly high doped Al-composition graded AlxGa1−xN films challenging for MOCVD. Equation (1) predicts that for a 50 nm linearly down-graded contact layer with x graded from 0.7 to 0, like previous reports, an activated dopant density needs to be higher than 8 × 1018 cm−3, throughout the contact layer, to achieve low-resistance films—a challenging problem for MOCVD due to the reasons mentioned above.22,23 This indicates that the contact layer resistance is sensitive to the dopant incorporation, and it is critical for the activated dopant density to be higher than the negative polarization charge density. Thus, to achieve lower contact resistance in MOCVD grown graded contact layers, it is necessary to reduce the negative polarization charge density in the graded contact layers, ρgraded, by reducing the factor, zPxz—gradient of the polarization charge.

Two different contact layer structures were investigated for this study—both grown on an n-type Al0.70Ga0.30N channel with a doping of 4.5 × 1018 cm−3. For sample A [Fig. 1(a)], the contact layer was graded from x = 0.7 to x = 0 over 50 nm, like previous reports, which requires [Si+] > 8 × 1018 cm−3 for polarization charge compensation [Fig. 1(b)].22,23 For sample B [Fig. 1(c)], the Al-composition was graded from x = 0.7 to x = 0.3 over 150 nm such that a lower [Si+] concentration can compensate the negative polarization charge density [Fig. 1(d)]. While a lower x at the top surface of the contact layer yields a lower metal-semiconductor resistance, the contact layer resistance will be higher due to a higher negative polarization charge density due to larger compositional grading and vice versa. A terminating Al-composition of x = 0.3 was thus chosen for the second case since a metal-semiconductor contact resistance of low-10−5 Ω cm2 is expected for Al0.3Ga0.7N, which is sufficient for a proof of concept demonstration.24 The thickness was chosen to be 150 nm, to reduce the negative polarization charge density even further so as to ensure a low resistance contact layer with a uniform ND = 3 × 1018 cm−3, which can be easily achieved for MOCVD for x < 0.8.23 

FIG. 1.

(a) Schematic and (b) energy band-diagram of the access region of sample A with a 50 nm thick linearly reverse-graded contact layer with x graded from 0.7 to 0 and [Si+] = 8 × 1018 cm−3 and (c) schematic and (d) energy band-diagram of the access region of sample B with a 150 nm thick linearly reverse-graded contact layer with x graded from 0.7 to 0.3 and [Si+] = 3 × 1018 cm−3.

FIG. 1.

(a) Schematic and (b) energy band-diagram of the access region of sample A with a 50 nm thick linearly reverse-graded contact layer with x graded from 0.7 to 0 and [Si+] = 8 × 1018 cm−3 and (c) schematic and (d) energy band-diagram of the access region of sample B with a 150 nm thick linearly reverse-graded contact layer with x graded from 0.7 to 0.3 and [Si+] = 3 × 1018 cm−3.

Close modal

The structures were grown using MOCVD on an AlN on a sapphire template with a 100 nm channel with a doping density of 4.5 × 1018 cm−3 in the channel and 1 × 1019 cm−3 in the graded contact layer. The source/drain contacts were formed by depositing the Ti/Al/Ni/Au metal stack using e-beam evaporation. The contacts on sample B were then annealed at 850 °C for 30 s using a rapid thermal annealing (RTA) system to form alloyed contacts, followed by inductively couple plasma-reactive ion etching (ICP-RIE) plasma etch to define device isolation mesas. The active areas of the devices were defined by selectively recessing the graded contact layer between the source and drain contacts. Over-recessing was performed to ensure complete removal of the graded contact layer such that 40 nm of the channel remained after the contact-layer recess.

Electrical characteristics were measured using an Agilent B1500A semiconductor device analyzer. Two-terminal IV on the fabricated samples showed that sample A had nonlinear IV characteristics, which indicates that the negative polarization charge in the contact layer has not been fully compensated. This indicates that the activated dopant density was lower than 8 × 1018 cm−3. The slow-graded contact scheme on sample B, however, shows linear IV characteristics, which indicates that the contact layer in sample B has indeed been compensated—as expected from the reduced polarization charge scheme. A comparison of two-terminal IV characteristics of the two samples is shown Fig. 2. Hall measurements were performed on ungated four-terminal van der Pauw (VDP) structures, and the sheet resistance of the channel was found to be 5.6 kΩ/◻ with a sheet carrier density of 1.8 × 1013 cm−2 and a mobility of 56 cm2/V s. Transfer length measurements (TLM) performed on gated TLM structures with a recessed contact layer on sample B yielded a specific contact resistivity of 3.3 × 10−5 Ω cm2. TLM and Hall measurements were not performed on sample A, due to the nonlinear nature of the contacts. A comparison of the specific contact resistivity for state-of-the-art MOCVD-grown AlxGa1−xN channel transistors with x > 0.5 is shown in Fig. 3. As can be seen, the specific contact resistivity obtained for sample B is the lowest observed for any MOCVD grown AlxGa1−xN channel devices to date for x > 0.5. The contact layer resistivity was estimated to be approximately 1 × 10−5 Ω cm2 which is, however, higher than that of MBE grown contact layers. Since Eq. (1) predicts that the incorporated dopant density should be sufficient to compensate for the negative polarization charge and provide a low resistance contact, the higher resistance could originate from (a) nonuniform grading of the contact layer leading to higher localized polarization charge and/or (b) lower localized [Si+] incorporation—both of which can lead to high resistance regions. This indicates that further growth optimization is required.

FIG. 2.

Comparison of two-terminal IV of samples A and B for two-terminal structures with LSD = 9 μm.

FIG. 2.

Comparison of two-terminal IV of samples A and B for two-terminal structures with LSD = 9 μm.

Close modal
FIG. 3.

Comparison of the specific contact resistivity vs Al-content for state-of-the-art MOCVD-grown AlxGa1−xN channel transistors with x > 0.5.5–7,9,10,22,23,26,27

FIG. 3.

Comparison of the specific contact resistivity vs Al-content for state-of-the-art MOCVD-grown AlxGa1−xN channel transistors with x > 0.5.5–7,9,10,22,23,26,27

Close modal

Metal-semiconductor field effect transistor (MESFET) structures were fabricated on sample B by depositing a gate metal stack of Ni/Au/Ni using e-beam evaporation. A schematic of the final fully processed sample structure is shown in Fig. 4. Transfer IV characteristics [Fig. 5(a)] measured for devices with LG = 0.6 μm and LSD = 1.5 μm (VDS = +20 V) showed a pinch-off voltage of −16 V and a maximum transconductance of 38 mS/mm. Output electrical characteristics [Fig. 5(b)] measured on the same device showed a maximum current density of 635 mA/mm (VGS = +2 V), which is the highest current density achieved to date for AlxGa1−xN channel devices with x > 0.5. 2-D technology computer aided design (TCAD) simulator, Silvaco, was used to model the described device structure.25 As can be seen, good match is obtained for transfer characteristics and the saturated current density. However, some deviation is observed in the slope of the output curves in the linear region—most likely originating from a mismatch between the field dependent mobility model in Silvaco and the actual device characteristics. Three-terminal breakdown characteristics of the MESFETs (Fig. 6) measured at VGS = −20 V for devices with a gate-drain spacing of LGD = 0.77 μm, showed no breakdown up to VGD = +220 V, which translates to an average field of 2.86 MV/cm, almost 3× higher than that exhibited by lateral GaN channel devices with similar dimensions. The breakdown is mainly limited by the gate leakage current which is the primary contributor to the drain current in the three terminal breakdown measurement. Thus, the breakdown characteristics can be further improved by the addition of a gate dielectric such as Al2O3.22 

FIG. 4.

Schematic of the fully processed sample B with a 150 nm reverse Al-composition graded contact layer with a channel thickness of 40 nm.

FIG. 4.

Schematic of the fully processed sample B with a 150 nm reverse Al-composition graded contact layer with a channel thickness of 40 nm.

Close modal
FIG. 5.

(a) Transfer characteristics and (b) family output I–V characteristics for device with a gate-length of LG = 0.6 μm and a source-drain spacing of LSD = 1.5 μm. Square symbols represent measured electrical characteristics, while solid lines represent simulated characteristics.

FIG. 5.

(a) Transfer characteristics and (b) family output I–V characteristics for device with a gate-length of LG = 0.6 μm and a source-drain spacing of LSD = 1.5 μm. Square symbols represent measured electrical characteristics, while solid lines represent simulated characteristics.

Close modal
FIG. 6.

Three-terminal breakdown voltage measurement showed no breakdown up to VGD = +220 V measured at VGS = −20 V for a gate-drain spacing, LGD = 0.77 μm.

FIG. 6.

Three-terminal breakdown voltage measurement showed no breakdown up to VGD = +220 V measured at VGS = −20 V for a gate-drain spacing, LGD = 0.77 μm.

Close modal

In summary, we have designed and demonstrated an improved MOCVD grown Al-composition graded contact layer with a contact resistance of 3.3 × 10−5 Ω cm2—lowest for the MOCVD grown AlxGa1−xN channel transistor with x > 0.5. Al0.7Ga0.3N channel transistors (LG = 0.6 μm and LSD = 1.5 μm) with this improved Al-composition graded contact layer exhibited a maximum current density of 635 mA/mm. Transistors with LGD = 0.77 μm did not undergo breakdown up to VGD = +220 V which corresponds to an average field of 2.86 MV/cm which is close to 3× higher than lateral GaN channel devices with similar dimensions. This demonstration provides a technologically important approach to form low resistance contacts to MOCVD grown ultrawide bandgap (UWBG) AlxGa1−xN channel transistors.

The authors acknowledge funding from the Air Force Office of Scientific Research (AFOSR Grant No. FA9550-17-1-0227, Program Manager Kenneth Goretta) and the DARPA DREaM program (No. ONR N00014-18-1-2033, Program Manager Dr. Young-Kai Chen, monitored by the Office of Naval Research, Program Manager Dr. Paul Maki).

1.
J. L.
Hudgins
,
G. S.
Simin
,
E.
Santi
, and
M. A.
Khan
,
IEEE Trans. Power Electron.
18
(
3
),
907
(
2003
).
2.
A. F. M.
Anwar
,
S.
Wu
, and
R. T.
Webster
,
IEEE Trans. Electron Devices
48
(
3
),
567
(
2001
).
3.
M.
Farahmand
,
C.
Garetto
,
E.
Bellotti
,
K. F.
Brennan
,
M.
Goano
,
E.
Ghillino
,
G.
Ghione
,
J. D.
Albrecht
, and
P. P.
Ruden
,
IEEE Trans. Electron Devices
48
(
3
),
535
(
2001
).
4.
T.
Razzak
,
H.
Xue
,
Z.
Xia
,
S.
Hwang
,
A.
Khan
,
W.
Lu
, and
S.
Rajan
, in
IEEE MTT-S International Microwave Workshop Series on Advanced Materials and Processes for RF and THz Applications (IMWS-AMP)
(
2018
).
5.
A. M.
Armstrong
,
B. A.
Klein
,
A.
Colon
,
A. A.
Allerman
,
E. A.
Douglas
,
A. G.
Baca
,
T. R.
Fortune
,
V. M.
Abate
,
S.
Bajaj
, and
S.
Rajan
,
Jpn. J. Appl. Phys., Part 1
57
(
7
),
074103
(
2018
).
6.
A. G.
Baca
,
A. M.
Armstrong
,
A. A.
Allerman
,
E. A.
Douglas
,
C. A.
Sanchez
,
M. P.
King
,
M. E.
Coltrin
,
T. R.
Fortune
, and
R. J.
Kaplar
,
Appl. Phys. Lett.
109
(
3
),
033509
(
2016
).
7.
S.
Muhtadi
,
S.
Hwang
,
A.
Coleman
,
F.
Asif
,
A.
Lunev
,
M. V. S.
Chandrashekhar
, and
A.
Khan
,
Appl. Phys. Lett.
110
(
17
),
171104
(
2017
).
8.
S.
Muhtadi
,
S. M.
Hwang
,
A.
Coleman
,
F.
Asif
, and
A.
Khan
,
in 75th Annual Device Research Conference (DRC)
(
2017
).
9.
S.
Muhtadi
,
S. M.
Hwang
,
A.
Coleman
,
F.
Asif
,
G.
Simin
,
M. V. S.
Chandrashekhar
, and
A.
Khan
,
IEEE Electron Device Lett.
38
(
7
),
914
(
2017
).
10.
N.
Yafune
,
S.
Hashimoto
,
K.
Akita
,
Y.
Yamamoto
,
H.
Tokuda
, and
M.
Kuzuhara
,
Electron. Lett.
50
(
3
),
211
(
2014
).
11.
S.
Bajaj
,
F.
Akyol
,
S.
Krishnamoorthy
,
Y.
Zhang
, and
S.
Rajan
,
Appl. Phys. Lett.
109
(
13
),
133508
(
2016
).
12.
S.
Bajaj
,
T.-H.
Hung
,
F.
Akyol
,
D.
Nath
, and
S.
Rajan
,
Appl. Phys. Lett.
105
(
26
),
263503
(
2014
).
13.
S.
Bajaj
, Ph.D. Thesis (
The Ohio State University
,
2018
).
14.
S.
Rajan
,
H.
Xing
,
S.
DenBaars
,
U. K.
Mishra
, and
D.
Jena
,
Appl. Phys. Lett.
84
(
9
),
1591
(
2004
).
15.
D.
Jena
,
S.
Heikman
,
D.
Green
,
D.
Buttari
,
R.
Coffie
,
H.
Xing
,
S.
Keller
,
S.
DenBaars
,
J. S.
Speck
, and
U. K.
Mishra
,
Appl. Phys. Lett.
81
(
23
),
4395
(
2002
).
16.
J.
Hwang
,
W. J.
Schaff
,
L. F.
Eastman
,
S. T.
Bradley
,
L. J.
Brillson
,
D. C.
Look
,
J.
Wu
,
W.
Walukiewicz
,
M.
Furis
, and
A. N.
Cartwright
,
Appl. Phys. Lett.
81
(
27
),
5192
(
2002
).
17.
B.
Borisov
,
V.
Kuryatkov
,
Y.
Kudryavtsev
,
R.
Asomoza
,
S.
Nikishin
,
D. Y.
Song
,
M.
Holtz
, and
H.
Temkin
,
Appl. Phys. Lett.
87
(
13
),
132106
(
2005
).
18.
Y.
Taniyasu
,
M.
Kasu
, and
N.
Kobayashi
,
Appl. Phys. Lett.
81
(
7
),
1255
(
2002
).
19.
P.
Pampili
and
P. J.
Parbrook
,
Mater. Sci. Semicond. Process.
62
,
180
(
2017
).
20.
F.
Mehnke
,
T.
Wernicke
,
H.
Pingel
,
C.
Kuhn
,
C.
Reich
,
V.
Kueller
,
A.
Knauer
,
M.
Lapeyrade
,
M.
Weyers
, and
M.
Kneissl
,
Appl. Phys. Lett.
103
(
21
),
212109
(
2013
).
21.
Y.-H.
Liang
and
E.
Towe
,
Appl. Phys. Rev.
5
(
1
),
011107
(
2018
).
22.
S.
Bajaj
,
A.
Allerman
,
A.
Armstrong
,
T.
Razzak
,
V.
Talesara
,
W.
Sun
,
S. H.
Sohel
,
Y.
Zhang
,
W.
Lu
, and
A. R.
Arehart
,
IEEE Electron Device Lett.
39
(
2
),
256
(
2018
).
23.
S.
Muhtadi
,
S.
Hwang
,
A.
Coleman
,
F.
Asif
,
A.
Lunev
,
M. V. S.
Chandrashekhar
, and
A.
Khan
,
Appl. Phys. Lett.
110
(
19
),
193501
(
2017
).
24.
B. A.
Klein
,
A. G.
Baca
,
A. M.
Armstrong
,
A. A.
Allerman
,
C. A.
Sanchez
,
E. A.
Douglas
,
M. H.
Crawford
,
M. A.
Miller
,
P. G.
Kotula
, and
T. R.
Fortune
,
ECS J. Solid State Sci. Technol.
6
(
11
),
S3067
(
2017
).
25.
Silvaco International
,
Device Simulation Software, ATLAS User's Manual
(
Silvaco International
,
Santa Clara, CA
,
2009
).
26.
E. A.
Douglas
,
S.
Reza
,
C.
Sanchez
,
D.
Koleske
,
A.
Allerman
,
B.
Klein
,
A. M.
Armstrong
,
R. J.
Kaplar
, and
A. G.
Baca
,
Phys. Status Solidi A
214
(
8
),
1600842
(
2017
).
27.
H.
Tokuda
,
M.
Hatano
,
N.
Yafune
,
S.
Hashimoto
,
K.
Akita
,
Y.
Yamamoto
, and
M.
Kuzuhara
,
Appl. Phys. Express
3
(
12
),
121003
(
2010
).