Download Hot-electron induced defect generation in AlGaN/GaN high electron

Solid-State Electronics xxx (2012) xxx–xxx
Contents lists available at SciVerse ScienceDirect
Solid-State Electronics
journal homepage: www.elsevier.com/locate/sse
Letter
Hot-electron induced defect generation in AlGaN/GaN high electron
mobility transistors
Hemant Rao ⇑, Gijs Bosman 1
Department of Electrical and Computer Engineering, 1064 Center Dr., 585 New Engineering Building, University of Florida, Gainesville, FL 32611, USA
a r t i c l e
i n f o
Article history:
Received 1 March 2012
Received in revised form 10 June 2012
Accepted 17 June 2012
Available online xxxx
a b s t r a c t
Hot-carrier effects in gated AlGaN/GaN based high electron mobility transistors are studied by microwave
noise temperature spectroscopy. Electron temperature profiles are measured and correlated to defect
generation in the AlGaN/GaN channel interface. It is found that new defects are created at the AlGaN/
GaN interface due to hot-electrons which are fairly well confined to the channel under the gated part.
Ó 2012 Elsevier Ltd. All rights reserved.
The review of this paper was arranged by S.
Cristoloveanu
Keywords:
Hot-electrons
Device reliability
Defect generation
1. Introduction
The high LO phonon energy (92 meV) and large inter-valley
separation (2 eV) in nitride semiconductors ensure significant
generation of hot-electrons during normal operation of AlGaN/
GaN based High Electron Mobility Transistors (HEMTs). Furthermore, a smaller LO phonon emission lifetime compared to LO phonon decay lifetime ensures substantial build-up of the lattice
temperature due to the strong electron–phonon coupling. In essence this creates a formidable combination of non-equilibrium
phonons which are energetic but highly localized due to their
low group velocity and hot-electrons which are mobile and energetic. These electrons have enough energy to degrade device performance temporarily and permanently [1]. Fig. 2 shows how
these electrons negatively impact the device performance by interaction with various parts of the device structure like the barrier,
buffer or the gate stack. Many experimental studies have demonstrated the existence of these hot-electrons in AlGaN/GaN channels
or GaN buffer materials [2–4]. Electron temperature profiles have
been measured directly in un-gated AlGaN/GaN channels by
microwave noise temperature spectroscopy [5]. Other groups have
measured electron temperature profiles using photolumiscence
spectra [6]. Very few experimental studies exist on gated AlGaN/
⇑ Corresponding author. Tel.: +1 352 392 4900; fax: +1 352 392 8381.
E-mail addresses: phemantrao@ufl.edu (H.
(G. Bosman).
1
Tel.: +1 352 392 0910; fax: +1 352 392 8381.
Rao),
[email protected]fl.edu
GaN HEMTs. Furthermore, realistic bias conditions have not been
probed. This work demonstrates the use of microwave noise spectroscopy to measure electron temperature profiles in a fully
processed advanced GaN HEMT. This is used to understand electrically stressed GaN HEMTs which underwent high channel electric
field stress and resulted in permanent defect generation at the AlGaN/GaN interface. Low frequency noise is used as a defect spectroscopic tool to confirm microscopic defect generation and their
spatial location in the device.
2. Experimental details
The devices under study are gated Al0.26Ga0.74N barrier based
GaN transistors grown on a silicon (1 1 1) substrate by MOCVD.
Schottky gate and source/drain ohmic contacts are made of Ni/Au
and Ti/Al/Ti/Au metal stack respectively formed by rapid thermal
annealing. They also have a source connected field-plate and a
SiN passivation layer. The devices were fully packaged preventing
any ambient light related instability to affect the measurements.
Fig. 1 shows the critical dimensions of the device. The microwave
device noise temperature is measured via Gasquet’s circulator
method [8] modified for a FET structure by the authors [9]. The
noise was measured at 2.20 GHz, beyond the spectral range where
the 1/f-like noise observed at low frequencies has a contribution.
The measurements were not performed under pulsed condition
thereby allowing the lattice temperature to reach steady state at
each bias. The gate terminal was AC-open circuited with the help
of a tuner to minimize the impact of both induced gate noise and
0038-1101/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.sse.2012.06.014
Please cite this article in press as: Rao H, Bosman G. Hot-electron induced defect generation in AlGaN/GaN high electron mobility transistors. Solid State
Electron (2012), http://dx.doi.org/10.1016/j.sse.2012.06.014
2
H. Rao, G. Bosman / Solid-State Electronics xxx (2012) xxx–xxx
Fig. 1. Crossection of devices used in this study.
thermal noise associated with gate resistance on the channel noise.
It was found that at high drain biases the drain noise mainly
stemmed from the channel electrons since the transconductance
of the device was relatively constant while the measured device
noise temperature increased with bias. The setup to measure low
frequency noise is detailed in a prior work [7]. The devices were
subjected to two different stress regimes where electric fields in
the channel were kept constant but the number of carriers were
varied by changing the gate voltage. VDS = 20 V and IDS = 230 mA
stress is a hot-carrier and self-heating stress. This is a fairly harsh
stress and the duration of the stress is 5 min. The second stress is
applied at VDS = 30 V and IDS = 1 mA for 60 min and generates hot
electrons without much self-heating.
3. Results and discussion
It was found that the hot-carrier and self-heating stress led to a
permanent increase of low frequency channel noise characteristics.
By measuring the normalized drain noise dependence on drain current it was found that the defect generation primarily occurred in
the gated part of the channel [10]. It is discussed in more detail later. This indicates that hot-electrons along with self-heating are
responsible for this effect. The other stress at low IDS 1 mA was
primarily hot-carriers and quite benign to the device. Only transient effects like threshold voltage increase and recovery were observed during and after stress. Low frequency channel noise
confirmed that no degradation occurred in the channel. To gain
further insight into this hot-carrier and self-heating induced failure
mechanism noise temperature measurements were performed. In
a previous paper [9] the authors measured noise temperatures as
a function of both drain and gate bias. It was demonstrated that
noise temperature is well correlated to the channel electric field
and not to the self-heating in the device. In this work the authors
analyze this data to correlate device noise temperatures to an
effective electron temperature in the channel.
In a FET biased in the linear region, the channel electric field is
uniform in the direction of the electron flow. Once the device goes
into saturation the electric fields are no longer uniform due to
pinch-off effects. A further increase of the drain voltage mainly
drops across a very narrow region of the channel near the drain
end and there the electric fields become two dimensional due to
a space charge region. For the linear region the electron temperature is equal to the noise temperature. In a previous work it was
observed that electron temperatures monotonically increase at
drain biases in the saturation region as well [9]. It is generally argued that the pinched-off region of the channel does not contribute
much to the measured channel noise since carriers are mostly traveling at the saturation velocity and they do not produce velocity
fluctuations [11]. Thus, the measured noise is mainly stemming
from the linear part of the channel where electric fields are still
uniform. From the DC ID–VD characteristics of the device it was
found that significant channel length modulation occurs after saturation. An empirical parameter (k) is often used to model the
channel length reduction due to drain bias. A value of 0.7 was extracted by extrapolating the output conductance in saturation for
different gate biases. The effective channel length then becomes
a function of drain bias
Leff ¼
LG
1 þ kV DS
ð1Þ
Here, LG is the gate length and VDS is the drain voltage above saturation. Using a simple MOSFET model the effective electric field in
the linear part of the channel is given by
ECH V GS V T
2Leff
ð2Þ
Fig. 3 shows the plot of noise temperature as a function of channel electric field for three different gate biases. Some interesting
features can be read from the data. Even though large electric fields
exist in the channel, the electron temperatures are not too high.
Fig. 2. Conduction band diagram along the gate to channel direction for a gate voltage higher than the threshold voltage. Equivalent electron temperatures in the channel are
plotted and two regions of high (15,000–25,000 K) and low energy (0–6000 K) electrons are shown. The effects of high and low energy electrons on DC and low frequency
noise characteristics are delineated.
Please cite this article in press as: Rao H, Bosman G. Hot-electron induced defect generation in AlGaN/GaN high electron mobility transistors. Solid State
Electron (2012), http://dx.doi.org/10.1016/j.sse.2012.06.014
H. Rao, G. Bosman / Solid-State Electronics xxx (2012) xxx–xxx
3
so-called hot-spots in the device but since they are farther away
from the channel they won’t affect the AlGaN/GaN interface much
and thus, the low frequency channel noise.
Fig. 4 shows the conduction band diagram along the gate to buffer direction simulated by the 1D Poisson–Schrodinger solver
NEXTNANO [14]. The equivalent electron temperature of the
2DEG is plotted in the channel. The measured electron temperatures fall within the low energy part of the channel between 0
and 3000 K. These electrons are fairly well confined along the AlGaN/GaN interface and have significant energies to generate new
defects at the interface. Due to built-in electric field in the AlGaN
barrier the threshold voltage of the device is more sensitive to
the charge distribution near the gate metal rather than the channel
interface. No change of threshold voltage or the gate leakage current was observed confirming that the electrons were restricted
in the channel.
Fig. 3. The effective electron temperature of the linear part of the channel of 0.5 um
is plotted as a function of channel electric field for three different gate biases
VGS = 1.2 (j), VGS = 1.1 (N) and VGS = 1.0 (d).
4. Conclusion
In summary, electron temperatures in the channel were extracted for a biased GaN HEMT as a function of channel electric
field for different gate biases. It was shown that hot-electrons are
generated in the gated part of the channel and are well confined
to the AlGaN/GaN interface. The high energy of these electrons
generates new defects at the AlGaN/GaN interface which is confirmed by channel low frequency noise measurements.
References
Fig. 4. Conduction band diagram along gate to channel region and the measured
electron temperatures are shown in the 0–3000 K band. This shows that electrons
are well confined at the AlGaN/GaN interface.
This could be due to thermal quenching of the hot-electrons since
measurements were performed at steady state without any application of pulsed techniques. This gives enough time for the
hot-electrons to lose their energy to the lattice. This trend is also
verified by two other factors. First, for low electric fields (<10 kV)
the electron temperatures are similar to as measured by other
groups [5,6]. Second, for higher electric fields the larger the gate
bias value, the lesser is the electron temperature. At higher gate
biases, more electrons exist in the channel for the same channel
electric field. This leads to larger self-heating and higher thermal
quenching of hot-electrons. Another interesting part is the physical
location of these hot-electrons in the channel. Due to channel
length reduction most of the hot-electrons are confined under
the gated part of the channel. Two dimensional simulations of
GaN HEMTs operating in the saturation region have shown that a
space charge zone forms in the pinched-off part of the channel
[12]. The electrons under this part are not confined to the quantum-well and move far away from the channel and deeper into
the buffer region [13]. Electric fields in that part are very high
and most of the electrons are highly energetic populating more
the upper valleys like L–M or C2. These electrons generate the
[1] Meneghesso G, Verzellesi G, Danesin F, Rampazzo F, Zanon F, Tazzoli A, et al.
Reliability of GaN high-electron-mobility transistors: state of the art and
perspectives. IEEE Trans Dev Mater Reliab 2008;8:332–43.
[2] Wang K, Simon J, Goel N, Jena D. Optical study of hot electron transport in GaN:
signatures of the hot-phonon effect. Applied Physics Letters 2006;88. 022103–
022103–3.
[3] Balkan N, Arikan MC, Gokden S, Tilak V, Schaff B, Shealy RJ. Energy and
momentum relaxation of hot electrons in GaN/AlGaN. Journal of Physics:
Condensed Matter 2002;14:3457–68.
[4] Matulionis A, Liberis J, Ardaravicius L, Ramonas M, Matulioniene I, Smart J.
Hot-electron energy relaxation time in AlGaN/GaN. Semiconductor Science
and Technology 2002;17:L9–L14.
[5] Matulionis A. GaN-based two-dimensional channels: hot-electron fluctuations
and dissipation. Journal of Physics: Condensed Matter 2009;21. 174203.
[6] Tripathy SK, Xu G, Mu X, Ding YJ, Wang K, Cao Y, et al. Evidence of hot electrons
generated from an AlN/GaN high electron mobility transistor. Applied Physics
Letters 2008;92. 013513–013513–3.
[7] Rao H, Bosman G. Simultaneous low-frequency noise characterization of gate
and drain currents in AlGaN/GaN high electron mobility transistors. Journal of
Applied Physics 2009;106:103712.
[8] Gasquet D, Vaissiere JC, Nougier JP. New method for wide band measurement
of noise temperature of one-port networks at high pulsed bias. In: sixth
international conference on noise in physical systems. NBS Special Publication;
1981. p. 305.
[9] Rao H, Bosman G. Device reliability study of GaN HEMTs using both low
frequency noise and microwave noise temperature spectroscopy. In: 2011 21st
International conference on noise and fluctuations (ICNF). IEEE; 2011. p.
464–7.
[10] Rao H, Bosman G. Device reliability study of AlGaN/GaN high electron mobility
transistors under high gate and channel electric fields via low frequency noise
spectroscopy. Microelectronics Reliability 2010;50:1528–31.
[11] Chen Chih-Hung, Deen MJ. Channel noise modeling of deep submicron
mosfets. IEEE Transactions on Electron Devices 2002;49:1484–7.
[12] Yin H. A physics-based large-signal analytical model for aluminum gallium
nitride/gallium nitride HFETs. North Carolina State University; 2008.
[13] Trew RJ. AlGaN/GaN HFET models and the prospects for physics-based
compact models. In: 2010 IEEE compound semiconductor integrated circuit
symposium (CSICS). IEEE; 2010. p. 1–4.
[14] Birner S, Zibold T, Andlauer T, Kubis T, Sabathil M, Trellakis A, et al. Nextnano:
general purpose 3-d simulations. IEEE Transactions on Electron Devices
2007;54:2137–42.
Please cite this article in press as: Rao H, Bosman G. Hot-electron induced defect generation in AlGaN/GaN high electron mobility transistors. Solid State
Electron (2012), http://dx.doi.org/10.1016/j.sse.2012.06.014