Solid-State Electronics 47 (2003) 10991104

www.elsevier.com/locate/sse

Low frequency noise in AlGaN/InGaN/GaN double

heterostructure eld eect transistors
N. Pala
a

a,b,*

, S. Rumyantsev a,1, M. Shur a, R. Gaska b, X. Hu b, J. Yang c,

G. Simin c, M.A. Khan c

Department of Electrical, Computer, and Systems Engineering, and Center for Integrated Electronics and Electronics Manufacturing,
CII 9017, Rensselaer Polytechnic Institute, Troy, NY 12180-3590, USA
b
Sensor Electronic Technology Inc., 1195 Atlas Road, Columbia, SC 29209, USA
c
Department of Electrical and Computer Engineering, University of South Carolina, Columbia, SC 29208, USA
Received 17 October 2002; accepted 19 November 2002

Abstract
Low-frequency noise in AlGaN/InGaN/GaN double heterostructure eld eect transistors was measured at room
and elevated temperatures as function of gate and drain voltages. Both 1=f noise and generationrecombination (gr)
noises were observed. The Hooge parameter, a, was estimated to be close to 1
103 . The activation energy for observed gr noise was found to be Ea  1:6 eV (the largest reported activation energy for GaN based devices). The
measurements also conrmed that the double heterostructure provided superior carrier connement in 2D channel even
at high carrier concentrations.
2003 Elsevier Science Ltd. All rights reserved.
Keywords: Low frequency noise; 1=f noise; Generationrecombination noise; Activation energy

1. Introduction
Group-III nitride heterostructure eld eect transistors (HFETs) have demonstrated impressive results in
high frequency and high power electronics. Recently
novel AlGaN/InGaN/GaN double heterostructure eldeect transistors (DHFETs) have been reported for high
power high frequency applications [1]. DHFETs particularly attracted attention since they showed no cur-

rent collapse in either the pulsed IV or the RF output

power characteristics.
Low frequency noise (LFN) is one of the major
factors determining the phase noise characteristics,
which are important for applications in microwave and
optical communication systems. The LFN properties of
GaN and AlGaN thin lms and AlGaN/GaN HFETs
have been studied in numerous papers [26].
In this article, we present experimental data on the
LFN in double heterostructure eld eect transistors
(DHFETs) and compare the results with those of LFN
studies for regular AlGaN/GaN HFETs.

Corresponding author. Address. Department of Electrical,

Computer, and Systems Engineering, and Center for Integrated
Electronics and Electronics Manufacturing, CII 9017, Rensselaer Polytechnic Institute, Troy, NY 12180-3590, USA. Tel.: +1518-276-6405; fax: +1-518-276-2990.
E-mail address: palan@rpi.edu (N. Pala).
1
On leave from Solid State Electronics Division, The Ioe
Physical-Technical Institute of Russian Academy of Sciences,
194021, St. Petersburg, Russia.

2. Experimental details
The double heterostructure employed in this study
was grown by low-pressure metal organic chemical vapor
deposition (LP-MOCVD) at 76 Torr and consisted of a
1.4 lm undoped GaN buer layer on i-SiC substrate,

0038-1101/03/$ - see front matter 2003 Elsevier Science Ltd. All rights reserved.
doi:10.1016/S0038-1101(02)00475-6

1100

N. Pala et al. / Solid-State Electronics 47 (2003) 10991104

 Inx Ga1x N layer with x < 0:1 and

followed by 50 A
 Al0:25 Ga0:75 N layer. Traces of indium
capped by a 250 A
were present through the structure and played an important role in improving the overall materials quality.
The growth temperatures for the GaN buer, InGaN
channel, and AlGaN barrier layers were 1000, 760, and
1100 C, respectively. Transistors with the gate length
and width of 1.5 and 200 lm and sourcedrain separation of 5 lm and transmission line model (TLM) test
)/Al(500 A
)/Ti(200
structures were fabricated. Ti(200 A
)/Au(1500 A
) metal layers were used for the ohmic
A
source and drain contacts. Ohmic contacts were annealed at 800 C for 60 s. Pt/Au Schottky contacts were
deposited as gate metal. The room temperature Hall
mobility and sheet carrier concentration for the deposited DHFET structures were measured to be 800 cm2 /
Vs and 1
1013 cm2 , respectively. This mobility value
was lower than typical values of the mobilities for regular AlGaN HFETs (typically around 10001500 cm2 /V s
with the record value of 2000 cm2 /V s [7]).
The LFN was measured in the frequency range from
1 Hz to 100 kHz in the temperature interval from 300 to
550 K in the common source conguration at small
values of the sourcedrain bias Vds (in the linear regime).
The voltage uctuations SV from the 100 X resistor
connected in series with the drain were analyzed by a
SR770 Network Analyzer.
The spectral noise density of the short circuit drain
current uctuations, SI , was calculated using the wellknown expression:


2
R rd
SI
SV
1
Rrd
where R 100 X is the load resistance, rd is the dierential resistance of the device under test. The temperature dependencies of noise were measured with the gate
grounded and at a constant drain voltage. The probe
station with 10-lm diameter tungsten probes and controlled pressure on the probes provided contacts to the
sample pads.

double heterostructure barrier layers extracted from the

capacitance measurements was 25 nm, in agreement with
the estimated value from the growth parameters.
Measurements of the currentvoltage characteristics
on the TLM structures with W 200 lm width and
distance, L, between the TLM pads ranging from 2 to 20
lm yielded the contact resistance of Rc 3 X mm and
the sheet resistance of Rsh 900 X/square.
The electron mobility ln in 2D gas can be extracted
from the measured IV characteristics:
ln

Lg
qns WRCh

where RCh Rtot  Rcs  Rs  Rd ; Rtot is the measured

drainsource resistance at low drain bias, Rcs is the total
series contact resistance Rcs 2Rc =W , Rs and Rd are the
series resistance of source-to-gate and gate-to-drain regions, respectively. The values Rc , Rs and Rd , were found
from the TLM measurements and known dimensions of
the transistor. The sheet concentration ns at the gate bias
VGS was found from capacitancevoltage measurements:
Z
1 VGS
ns
C dVgs
3
q VG0
where C is the capacitance per unit area, VG0 is the gate
voltage, which is taken to be well below the threshold
voltage, VT . Since at Vgs < VT , the gate capacitance is
very small, the chosen value of Vg1 does not signicantly
aect the dependence ns VGS given by Eq. (3) [8].
Fig. 1 shows the dependence of the electron mobility
ln in 2D-gas on channel electron sheet concentration ns .
The inset shows the same dependence for SiO2 /AlGaN/
GaN MOSHFETs reported in Ref. [9]. As seen in the

3. Results and discussion

3.1. The DC measurements
The capacitancevoltage measurements were performed at frequencies from 1 to 10 kHz on the test
Schottky diodes with the area of 4
104 cm2 in order
to nd thickness and doping level of the double heterostructure layers. For the entire range of the gate biases, Vg , (from zero to the threshold voltage, VT ) the gate
capacitance was practically constant, manifesting the
linear dependence of the electron channel sheet concentration ns on the gate bias. The thickness of the

Fig. 1. Electron mobility dependence on sheet concentration of

2D gas in channel of DHFET. The inset shows the reported
characteristics for SiO2 /AlGaN/GaN MOSHFETs in Ref. [8].

N. Pala et al. / Solid-State Electronics 47 (2003) 10991104

gure, in both devices, ln rst increases with ns , reaches

a maximum value, and then rapidly decreases with a
further increase of ns . However the concentration for the
maximum mobility in DHFET is ns 1:1
1013 cm2
which is almost the twice of that reported for MOSHFET. The increase of ln with ns can be explained by
increased screening of ionized impurities and dislocations in the 2D electron gas [10,11]. The subsequent
decrease of ln can be attributed to the electrons spillover from the 2D-channel to a parallel low electronmobility parasitic conduction channel [1214]. The
higher concentration value corresponding to the maximum mobility in DHFETs is evidence of the superior
carrier connement in the channel of these devices.
3.2. Noise measurements
Fig. 2 shows the noise spectrum for a device with
W 150 lm at room temperature. As seen, 1=f noise is
the dominant noise at the temperature of 300 K. The
1=f noise level in dierent semiconductor materials and
structures is usually characterized by the dimensionless
Hooge parameter, a [15]:
a

SI
fN
I2

where N is the total number of the conduction electrons

in the sample, f is the frequency of the analysis, SI =I 2 is
the relative spectral density of noise. Using the value of
N extracted from capacitancevoltage measurements, we
found a  1
103 . This value of a is comparable with
the values reported for regular AlGaN/GaN HFETs.
Contribution of the contact noise to the measured
noise spectra can be determined by the noise measurements on the TLM structures [16].
Assuming that the noise sources are not correlated
and located in the contacts and in the channel, the
spectral noise density of the current uctuations can be
expressed as follows:

Fig. 2. Noise spectra of relative drain current uctuations of

DHFET with W 150 lm L 1 lm.

1101

SId SRChTLM
R2ChTLM
SRc
2

Id2
RChTLM RChTLM RC 2 R2C

R2C
RChTLM RC 2

where SRChTLM , and SRc are spectral densities of channel

resistance RChTLM between contact pads, and contact
resistance RC uctuations, respectively. When RChTLM
RC , the spectral noise density, SI =I 2 , should be proportional to 1=L2 if the contact noise is dominant
(SRc SRChTLM ) and to 1=L if the spectral noise density
of the channel resistance uctuations are dominant
(SRc
SRChTLM ).
Symbols in Fig. 3 show experimental dependence of
noise measured at frequency f 100 Hz as a function of
the distance between contact pads in the TLM structure.
Since this dependence is close to the 1=L law, we conclude that contacts do not contribute much to the
overall noise.
The concentration dependence of noise in AlGaN/
GaN HFETs and SiO2 /AlGaN/GaN MOSHFETs has
been discussed in the past [8,17]. To assess the concentration dependence of noise, a more accurate procedure
of the a extraction should take into account dierent
noise properties of the channel and source-to-gate, gateto-drain regions, as well as the eect of the series drain
and source contact resistances. To account for this difference, we introduce two Hooge parameters: the Hooge
parameter for the channel under the gate aCh and the
Hooge parameter for the ungated regions a0 . We estimated these parameters using the procedures developed
in Ref. [17].
Assuming that the noise sources in gated and ungated
parts of the transistor are not correlated the dependence
of aCh on the channel concentration ns is given by

Fig. 3. The dependence of the relative spectral noise density

SI =I 2 on the distance L between the pads of TLM structures.
f 100 Hz.

1102

N. Pala et al. / Solid-State Electronics 47 (2003) 10991104


aCh


SI
a0 R2sd
ACh ns R2tot

Id2 Asd nsd R2tot
R2Ch

where RCh and Rsd Rs Rd are resistances of the

gated and ungated regions, respectively, Asd is the total
area of the drain-to-gate, gate-to-source openings, nsd is
the electron concentration in these regions, ns is the
electron concentration in the channel given by the Eq.
(3). At Vg 0 the 2D concentrations under the gate and
in ungated regions do not dier much. Therefore the
value of a0 can be estimated assuming that the value of
the Hooge parameter is the same for the channel and
ungated regions at zero gate voltage.
Fig. 4 shows the dependence of the Hooge parameter
aCh on the 2D sheet channel concentration, ns , calculated
for the DHFETs. At low channel concentrations, aCh
decreases with the increase of ns as aCh  1=ns , reaches a
minimum and then increases with a further increase of
ns . The minimum of the Hooge parameter in Fig. 4
corresponds to nearly the same concentration ns as the
maximum in the mobility curve shown in Fig. 1.
The dependence aCh  1=ns is often observed in
n-channel Si MOSFETs [18,19] as a result of tunneling
of electrons from semiconductor to traps in the oxide.
The same dependence was recently observed in AlGaN/
GaN HFETs [8] and explained by the tunneling from the
2D gas into the traps in AlGaN or/and GaN layers.
Since we observed the same behavior in AlGaN/InGaN/
GaN (DHFETs), the measured noise also might be
linked to the tunneling mechanism.
There might be several reasons for the increase of the
aCh at high concentration ns . As discussed above, the
decrease of the mobility at high concentrations, ns , can
be attributed to the electron spillover from the 2Dchannel to a parallel low electron-mobility parasitic
conduction channel. The noise level in such parasitic
channel could be very large and corresponds to
a  102 1 for the bulk GaN [20,21]. Another reasons

Fig. 5. (a)Temperature dependence of LFN. (b) Arrhenius plot

for LFN. Slope gives the activation energy as 1.6 eV.

for aCh increase are contribution for noise from the

forward gate current and from the ungated parts of the
transistor which might become dominant when channel
noise is very small.
Fig. 5a shows the temperature dependence of the
noise spectral density SI =I 2 in the frequency range from
100 to 3000 Hz. The temperature dependence of the
measured noise spectral density has a wide and pronounced maximum at elevated temperatures. The temperature Tmax corresponding to the maximum noise
increases with frequency that is typical for the generationrecombination noise caused by a local level [2224].
Fig. 5b shows the dependence of 1=kTmax versus lnf
(the Arrhenius plot). The slope of this dependence gives
the activation energy Ea 1:6 eV. This is the largest
reported activation energy for GaN based devices.

4. Conclusions
Fig. 4. The dependence of Hooge parameter aCh on 2D sheet
carrier concentration ns for DHFETs.

LFN in AlGaN/InGaN/GaN (DHFETs was studied

within the temperature range from 300 to 520 K. Both,

N. Pala et al. / Solid-State Electronics 47 (2003) 10991104

1=f and generation recombination noise were found.

The Hooge parameter, a, was estimated to be about
1
103 . Noise measurements on TLM structures
showed that contacts do not contribute much to the
noise. The activation energy for observed gr noise was
found to be Ea  1:6 eV. This is the largest reported
activation energy for GaN based devices. The concentration dependence of mobility showed that the double
heterostructure provided superior carrier connement in
2D channel at high carrier concentrations. This result
might be linked to a smaller energy gap in the InGaN
quantum well layer (see recent data on small energy gap
of InN [25,26]). At low channel concentrations, aCh decreases with the increasing channel concentration following 1=ns dependence. This dependence might be the
evidence of noise generated by electron tunneling to
AlGaN barrier layer or to GaN bulk layer.

[8]

[9]

[10]

[11]
[12]

[13]

[14]

Acknowledgements
The work at RPI was supported by the Oce of
Naval Research, and the project was monitored by H.
Dietrich. The work at SET, Inc. was supported by the
Missile Defense Agency under contract DASG60-00-C0088 and monitored by F. Clarke of the Army/SMDC.
The work at USC was supported partially by MDA
under Army SMDC contract DASG60-10003, monitored by Mr. T. Bauer, and by DARPA contract DAA
D19-02-10236 monitored by Dr. E. Martinez and Dr. A.
Hung.

[15]

[16]

[17]

References
[1] Simin G, Hu X, Tarakji A, Zhang J, Koudymov A, Saygi
S, et al. AlGaN/InGaN/GaN double heterostructure eldeect-transistors. Jpn J Appl Phys 2001;40:L11424.
[2] Pala N, Rumyantsev SL, Shur MS, Gaska R, Hu X, Yang
J, et al. Generationrecombination and 1=f noise in
Al0:4 Ga0:6 N thin lms. Fluctuation Noise Lett, in press.
[3] Balandin A. Gate voltage dependence of low frequency
noise in GaN/AlGaN heterostructure eld eect transistors. Electron Lett 2000;36(10):9123.
[4] Rumyantsev SL, Pala N, Shur MS, Gaska R, Levinshtein
ME, Asif Khan M, et al. Thin n-GaN lms with low level
of the 1=f noise. Electron Lett 2001;37(11):720.
[5] Pala N, Gaska R, Rumyantsev S, Shur MS, Asif Khan M,
Hu X, et al. Low-frequency noise in AlGaN/GaN MOSHFETs. Electron Lett 2000;36(3):268.
[6] Rumyantsev SL, Pala N, Shur MS, Borovitskaya E,
Dmitriev AP, Levinshtein ME, et al. Generationrecombination noise in GaN/GaAlN heterostructure eld eect
transistors. IEEE Trans Electron Dev 2001;48(3):530.
[7] Gaska R, Yang JW, Osinsky A, Chen Q, Asif Khan M,
Orlov AO, et al. Electron transport in AlGaNGaN

[18]

[19]

[20]

[21]

[22]

[23]

1103

heterostructures grown on 6H-SiC substrates. Appl Phys

Lett 1998;72(6):7079.
Rumyantsev SL, Pala N, Shur MS, Gaska R, Levinshtein
ME, Ivanov PA, et al. Concentration dependence of the
1=f noise in AlGaN/GaN heterostructure eld eect
transistors. Semicond Sci Technol 2002;17:4769.
Ivanov PA, Levinshtein ME, Simin G, Hu X, Yang J, Asif
Khan M, et al. Drift mobility of electrons in AlGaN/GaN
MOSHFET. Electron Lett 2001;37(24):147981.
Weimann NG, Eastman LF, Doppalapudi D, Ng HM,
Maustakus TD. Scattering of electrons at threading dislocations in GaN. J Appl Phys 1998;83(7):36569.
Ando T, Fowler AB, Stern F. Electronic properties of twodimensional systems. Rev Modern Phys 1982;54(2).
Gaska R, Shur MS, Bykhovski AD, Orlov AO, Snider GL.
Electron mobility in modulation doped AlGaNGaN
heterostructures. Appl Phys Lett 1999;74(2):2879.
Dang XZ, Asbeck PM, Yu ET, Sullivan GJ, Chen MY,
McDermott BT, et al. Measurement of drift mobility in
AlGaN/GaN heterostructure eld-eect transistor. Appl
Phys Lett 1999;74(25):38902.
Knap W, Borovitskaya E, Shur MS, Gaska R, Karczewski, G, Brandt B, et al. High magnetic eld studies of
AlGaN/GaN heterostructures grown on bulk GaN,
SiC, and sapphire substrates. In: Wetzel G, Shur MS,
Mishra UK, Gil B, Kishino K, editors. Materials Research Society Symposium Proceedings, Vol. 639, G7.3.
2001.
Hooge FN, Kleinpenning TGM, Vandamme LKJ. Experimental studies on 1=f noise. Rep Progr Phys 1981;44(5):
479532.
Rumyantsev SL, Pala N, Shur MS, Gaska R, Levinshtein
ME, Asif Khan M, et al. Low frequency noise in GaN
metal semiconductor and metal oxide semiconductor eld
eect transistors. J Appl Phys 2001;90(1):3104.
Rumyantsev SL, Pala N, Shur MS, Levinshtein ME,
Ivanov PA, Asif Khan M, et al. Low-frequency noise
in AlGaN/GaN heterostructure eld eect transistors
and metal oxide semiconductor heterostructure eld effect transistors. Fluctuation Noise Lett 2001;1(4):L2216.
McWhorter AL. 1=f Noise and germanium surface properties. In: Kingston RH, editor. Semiconductor surface
physics. Philadelphia PA: University of Pennsylvania
Press; 1957. p. 20728.
Vandamme LKJ, Li X, Rigaud D. 1=f noise inMOS
devices mobility or number uctuations. IEEE Trans
Electron Dev 1994;41(11):193645.
Dyakonova NV, Levinshtein ME, Contreras S, Knap W,
Beaumont B, Gibart P. Low-frequency noise in n-GaN.
Semiconductors 1998;32:257.
Levinshtein ME, Rumyantsev SL, Look DC, Molnar RJ,
Asif Khan M, Simin G, et al. Low-frequency noise in nGaN with high electron mobility. J Appl Phys
1999;86:5075.
Sodini D, Touboul A, Lecoy G, Savelli M. Generation
recombination noise in the channel of GaAs Schottky-gate
eld-eect transistors. Electron Lett 1976;12:423.
Scholz FJ, Hwang JM, Schroeder DK. Low frequency noise and DLTS as semiconductor device characterization tools. Solid State Electron 1988;31(2):
20517.

1104

N. Pala et al. / Solid-State Electronics 47 (2003) 10991104

[24] Levinshtein ME, Rumyantsev SL. Noise spectroscopy of

local levels in semiconductors. Semicond Sci Technol
1994;9(6):11839.
[25] Davydov VYu, Klochikhin AA, Emtsev VV, Ivanov SV,
Vekshin VV, Bechstedt F, et al. Band gap of InN and In-

rich Inx Ga1x N alloys (0:36 < x < 1). Physica Status Solidi
B 2002;230(2):R46.
[26] Wu J, Walukiewicz W, Yu KM, Ager III JW, Haller EE,
Lu Hai, et al. Unusual properties of the fundamental band
gap of InN. Appl Phys Lett 2002;80(21):39679.