Download Low-Frequency Noise and Interface States in GaAs Homojunction

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 46, NO. 4, APRIL 1999
IV. CONCLUSION
A new low-temperature photoluminescence technique was shown
to be effective in identifying Be out-diffusion in AlGaAs/GaAs
HBT’s. The degree of Be out-diffusion can be monitored simply by
measuring the energy separation between the two PL emission peaks
in AlGaAs range. This approach has been verified by comparing
the PL spectra that were obtained on AlGaAs/GaAs HBT structures
with different MBE growth conditions. In addition, dc parameters on
fabricated HBT’s such as dc gain and emitter-base turn-on voltage
show good correlation with the PL measurements.
ACKNOWLEDGMENT
The authors wish to thank Dr. S. F. Yoon, Dr. Rusli, Dr. B. S.
Ooi, C. L. Tang, K. S. Ang, and H. Yang for their valuable help and
discussion.
811
Low-Frequency Noise and Interface States in
GaAs Homojunction Far-Infrared Detectors
W. Z. Shen and A. G. U. Perera
Abstract—Low-frequency noise characteristics of p-GaAs homojunction
interfacial work-function internal photoemission (HIWIP) far-infrared
(FIR) detectors are reported. The noise was found to exhibit 1/f behavior
related to interface states at frequencies below 1 kHz and frequency
independent shot noise at higher frequencies. The noise expressions
correctly predict the dark current noise behavior, and provide a means
of estimating both the gain and energy distribution of the interface states.
The interface state density is estimated to be in the order of 1011 cm02 .
It has been shown that the estimated gain and noise equivalent power
are in good agreement with the previous results obtained via optical
measurements.
Index Terms—FIR detectors, GaAs, interface states, noise.
I. INTRODUCTION
REFERENCES
[1] K. Yang, J. R. East, and G. I. Haddad, “Numerical study on the injection
performance of AlGaAs/GaAs abrupt emitter heterojunction bipolar
transistors,” IEEE Trans. Electron Devices, vol. 41, p. 138, Feb. 1994.
[2] D. C. Streit, A. K. Oki, T. R. Block, M. D. Lammert, M. M. Hoppe, D.
K. Umemoto, and M. Wojtowicz, “Commercial heterojunction bipolar
transistor production by molecular beam epitaxy,” J. Vac. Sci. Technol.
B, vol. 14, p. 2216, 1996.
[3] J. N. Miller, D. M. Collins, and N. J. Moll, “Control of Be diffusion
in molecular beam epitaxy GaAs,” Appl. Phys. Lett., vol. 46, no. 10,
p. 960, 1985.
[4] J. I. Pankove, Optical Process in Semiconductors. Englewood Cliffs,
NJ: Prentice-Hall, 1971.
[5] H. B. Bebb and E. W. Williams, in R. K. Willardson and A. C. Beer,
Eds., Semiconductors and Semimetals. New York: Academic, vol. 8,
1972.
[6] T. Humer-Hager and H. Tews, “Photoluminescence in GaAs/AlGaAs
heterojunction bipolar transistors: Investigation of the properties of the
Mg acceptor,” J. Appl. Phys., vol. 68, no. 3, p. 1310, 1990.
[7] Z. H. Lu, M. C. Hanna, and A. Majerfeld, “Determination of band
gap narrowing and hole density for heavily C-doped GaAs by photoluminescence spectroscopy,” Appl. Phys. Lett., vol. 64, no. 1, p. 88,
1994.
[8] Z. H. Lu, A. Majerfeld, P. D. Wright, and L. W. Yang, “A comprehensive
optical characterization method for high-performance npn AlGaAs/GaAs
heterojunction bipolar transistors,” IEEE J. Select. Topics Quantum
Electron., vol. 1, p. 1030, 1995.
[9] G. W. Wang, R. L. Pierson, P. M. Asbeck, K. Wang, N. Wang, R.
Nubling, M. F. Chang, J. Salerno, and S. Sastry, “High-performance
MOCVD-grown AlGaAs/GaAs heterojunction bipolar transistor with
carbon-doped base,” IEEE Electron Device Lett., vol. 12, p. 347, June
1991.
High-performance far infrared (40 200 m) semiconductor
detectors as well as large focal plane arrays are required for space
astronomy applications, such as NASA’s airborne mission, stratospheric observation for infrared astronomy (SOFIA), and the ESA’s
far-infrared and sub-millimeter telescope (FIRST) programs. Si and
GaAs homojunction interfacial work-function internal photoemission
(HIWIP) far-infrared (FIR) detectors [1] can be strong competitors
for present extrinsic Ge photoconductors (unstressed or stressed [2])
and Ge block-impurity-band (BIB) detectors [3] due to the material
advantages. Significant progress has already been achieved in the
development of p-GaAs HIWIP FIR detectors [4], resulting
in a
p
responsivity of 3.1 A/W, detectivity of 5:9 2 1010 cm Hz/W, and
cutoff wavelength as long as 100 m.
Noise measurements provide a valuable diagnostic tool for the
evaluation of electronic and optoelectronic devices and their long term
performance. Characterization of low-frequency noise in HIWIP FIR
detectors is useful not only for improving the device performance,
but also for getting information about the physical properties, such
as interface states, gain, etc. However, the noise properties of HIWIP
FIR detectors have not been reported before. The aim of this article
is to present a systematic evaluation of the low-frequency noise in
p-GaAs HIWIP detectors.
The p-GaAs detector sample (no. 9604) studied here as reported
before [4] was grown by molecular beam epitaxy (MBE) with
epilayers consisting of a 3000-Å bottom contact (p++ )-layer, a 1500Å undoped (i)-layer, 20 periods of thin emitter (p+ )-layers (thickness
150 Å), undoped i-layers (thickness 800 Å), and, finally, a 3000-Å
top emitter layer and a 3000-Å top contact layer. The emitter layers
were doped with Be to 4 2 1018 cm03 . The detector [4], [5] mesa
area is 400 2 400 m2 and the optical window area is 260 2 260 m2 .
The noise characteristics were measured using a low-noise preamplifier (SR 560) and a fast Fourier transform (FFT) spectrum analyzer
(SR 780) with the detector temperature at 4.2 K. The equipment
was calibrated by measuring the room temperature noise level of
a conventional 4.6 k
resistor. Typical current noise spectra of the
Manuscript received September 2, 1998; revised November 18, 1998. The
review of this brief was arranged by Editor J. N. Hollenhorst. This work was
supported in part by the National Aeronautics and Space Administration under
Contract NAG5-4950.
The authors are with the Department of Physics and Astronomy, Georgia
State University, Atlanta, GA 30303 USA.
Publisher Item Identifier S 0018-9383(99)03592-3.
0018–9383/99$10.00  1999 IEEE
812
IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 46, NO. 4, APRIL 1999
Fig. 1. Measured dark current noise spectra of p-GaAs HIWIP far-infrared
detector at 4.2 K for various forward biases. The dashed line represents the
1/f dependence of the noise power density Si . The inset shows the spectral
response of the p-GaAs HIWIP FIR detector measured at 4.2 K under different
forward bias values. The deep valley at 36.5 m is due to the transverse optical
(TO) phonons of GaAs. The structures marked with arrows are believed to be
related with interface states. The hydrogenic excited states of the Be acceptors
are expected to show structures between 44 and 60 m.
(1)
stronger with increasing bias and can be attributed to the localized
nature of interface states [5]. The hydrogenic transitions (from 1s
to 2p, 3p, 4p, and continuum) of Be acceptor impurity in GaAs are
expected at 59.1, 49.8, 47.2, and 44.3 m, respectively. Here the
peak at 59.0 m which is between 44 and 60 m could be the
1s ! 2p transition. The local vibrational mode (LVM) absorption
of Be impurity in GaAs was reported at 20.7 m in literature
corresponding to the 100% abundant 9 BeGa acceptors [10].
The origin of the interface states, which normally have a sheet
density of 1010 –1012 cm02 , can be the dangling bonds in the
interfaces, Coulomb potential of charged ions, and impurities near
interfaces [11]. If the G-R current noise is mostly generated by
interface states near the Fermi level, the interface states can be
estimated from the noise power density (Si ) by the following equation
[9]:
=
2
Si (f ) = C Id
A0 fNis
studied p-GaAs HIWIP FIR detector at 4.2 K for various forward bias
values are presented in Fig. 1. Similar noise behavior was observed
under reverse bias conditions. Also observed is the symmetry in
dark current noise under forward and reverse biases due to the
symmetric current–voltage (I –V ) characteristics in HIWIP detectors
[5]. All the spectra display 1/f noise dependence at frequencies (f )
below 1 kHz and are independent of frequency at higher values. The
observed current noise spectra result from 1/f flicker noise and shot
noise spectra. Absence of Lorentzian-type noise in the noise spectra
indicates that the current noise power density can be written as
+ C II
is the amplitude of the flicker (1/f ) noise, and C II
qId g denotes the shot noise spectrum, as in quantum well infrared
photoconductors (QWIP’s) [6], with q as elementary charge, Id as
the detector forward dark current and g as noise gain of the detector.
At low frequencies (f kHz), the value of is found to vary
from 1.0 to 1.2 and no simple relationship was found between and bias. In order to understand the origin of the 1/f noise, a plot
of 1/f noise power density (Si ) as a function of dark current Id
at frequencies of 10, 100, and 500 Hz is measured and shown in
Fig. 2. It is found that the 1/f noise power density is proportional
to Id with an value of 2.05–2.10. This type of behavior indicates
that the origin of the 1/f noise could be interpreted in terms of a
random fluctuation in the occupancy of the interface trap centers
which can lead to generation-recombination (G-R) 1/f noise [9].
Two representative models of the 1/f noise have been proposed
[7], [8]. One model, the number fluctuation theory [7], is based on
the slow fluctuations in the total number of carriers taking part in
hopping conduction, resulting in the f 01 frequency and Id2 dark
current dependences of the noise power density. The other model [8]
predicts a linear dependence of Si on Id . The free carrier absorption
and internal photoemission in HIWIP detectors lead to carrier number
fluctuations, which would result in current fluctuations in the external
circuit when a net current flows through the detector. This kind of
noise is related to the presence of interface localized states [9], which
is in good agreement with the results from the detector response
spectra shown in the inset of Fig. 1, where remarkably reproducible
spike responses (marked with arrows) were observed. They become
where
4
CI
Si (f ) = C I Id
f
1
Fig. 2. 1/f noise power density Si as a function of the dark current Id at
frequencies of 10, 100, 500 Hz. The dashed line represents the Id2 dependence
of the noise power density Si .
(2)
where C is a constant which in practice 0.1, A0 is the detector area
( : 2 03 cm2 ), and Nis is the interface state density.
Since the energy distribution of interface states is determined
by the Fermi level, the density of interface states should change
exponentially with the Fermi level. This provides an independent
confirmation for the above explanation. A strong bias dependence
of spectral response, both responsivity and cutoff wavelength, was
observed and well explained for HIWIP structures [4], as a result of
the barrier lowering due to image force effect [1]. Under different
biases, the position of the Fermi level at the interfaces with respect
to the barrier is determined [1] by the barrier lowering
1 6 10
18 = 4qF0 s
18
(3)
where F is the electric field across the detector and 0 and s are
vacuum permittivity and relative dielectric constants, respectively.
Fig. 3 shows the density of interface states obtained from (2) as a
function of barrier lowering
calculated from (3). Accordingly
Nis was found to increase exponentially with the barrier lowering,
which can be well fitted (the solid line in Fig. 3) by an empirical
state-density distribution
18
exp 0 18
(4)
Ei
with N0 = 3:5 2 1011 cm02 and Ei = 1:753 meV. The estimated
11
02
Nis = N0
Nis is in the order of 10
cm
, a value which compared favorably
IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 46, NO. 4, APRIL 1999
813
18
Fig. 3. Interface state density Nis as a function of barrier lowering
due
to the image force effect, which changes the position of the Fermi level at the
interfaces with respect to the barrier. The solid line is a curve fitted to (4) with
11 cm02 , and E
fitting parameters of N0
:
:
meV. Shown
i
in the inset is the interface state density obtained via C –V measurements and
noise measurements.
= 3 5 2 10
= 1 753
with the density of interface states (2:5 2 1011 cm02 ) reported for
MBE grown Be-doped p-type GaAs [12].
The capacitance and conductance of the detector were also measured at 4.2 K with the help of a Hewlett-Packard multifrequency
LCR meter (4284 A). The interface state density was calculated and
integrated as a function of bias from the low-frequency (10 KHz)
2
capacitance data with the relation Nis =
Cis /q A0 , where Cis
was the interface state capacitance calculated from an equivalent
circuit of the detector, which is similar to the GaAs/AlGaAs quantum
well infrared photodetectors [13]. Shown in the inset of Fig. 3 is the
interface state density obtained via capacitance–voltage (C –V ) measurements, comparing with the results from the noise measurements.
The interface state results from C –V measurements roughly agree
with the order of 1011 cm02 and show the same bias dependent
behavior as the noise measurements. The C –V measurements are
much noisier at frequencies below 10 KHz. However, the estimation
from noise measurements is only valid below 1 KHz. This difference
in frequency may have resulted in the deviation in estimating the
interface states.
The noise measurements also provide a means for gain determination [6]. For frequencies above 1 kHz, the noise was independent
of frequency and was dominated by shot noise. The gain g can
be obtained using the current shot noise expression [C II in (1)].
Combining Si and Id allows the experimental determination of g
as shown in Fig. 4, where the smooth curve is drawn through the
experimental points. For this HIWIP detector, the determined gain
increases rapidly with bias at low voltages, and gradually saturates
(near the bias with the highest detectivity [4]). This behavior is similar
to the case of QWIP’s [6]. The highest value of g is 0.95 at a bias
corresponding to the highest responsivity [4]. This value of gain is in
good agreement with the previous estimation of 0.984, obtained by
combining the experimental responsivity and quantum efficiency [4].
Furthermore, using the optical gain equation, g = 1 0 p, derived
for HIWIP detectors [14], where p is the carrier trapping probability,
the shot noise power density can be rewritten, ignoring the difference
between noise gain and optical gain, as
Si
= 4qId (1
0
p):
The dark current in HIWIP detector is also related to
Id
=
qGA0 d(1
0
p)
Fig. 4. Experimentally determined gain g versus forward bias for the p-GaAs
HIWIP detector at 4.2K. The smooth curve is drawn through the measured
data. The inset shows the shot noise power density Si as a function of the
dark current Id at a frequency of 1500 Hz.
to escape. Substituting (6) into (5), we get
Si
=
2
4Id
GA0 d
(7)
which shows the shot noise power density is also proportional to Id2 ,
same as the 1/f noise case. This can be clearly seen in the inset
of Fig. 4, where the shot noise Si is displayed as a function of dark
current at a frequency of 1500 Hz. Similar results have been obtained
in GaAs/AlGaAs QWIP’s [15]. Furthermore, equaling (2) and (7)
yields for the corner frequency fc
fc
=
C Gd
4Nis
:
(8)
The fact that in Fig. 1 the corner frequency increases with increasing
bias suggests indeed that Nis is reducing with bias (provided G is
constant). However, the corner frequency variation with bias can not
be determined accurately (see Fig. 1), hence, Nis versus bias was
estimated via 1/f noise and C –V measurements.
The measured shot noise data can be used to directly estimate the
noise equivalent
power (NEP) in the p-GaAs HIWIP FIR detector via
p
NEP = Si =R, where R is the responsivity. At a bias of 89 mV, the
measured shot noise Si is 8:3 2 10025 A2 /Hz, and the responsivity
of the detector
p at that bias is 2.12 A/W. This yields pa NEP of 4.3
2 10013 W/ Hz (detectivity D3 of 6:0 2 1010 cm Hz/W), 3also
in good agreement
p with the previous optical estimation [4] D of
10
5:9 2 10
cm Hz/W at a bias of 83 mV.
In summary, the low-frequency noise in p-GaAs HIWIP farinfrared detectors has been investigated in detail. Below 1 kHz, the
observed 1/f noise can be well explained by the number fluctuation
theory, which has been employed to estimate the energy distribution
of the interface states. The noise was dominated by shot noise at
higher frequencies. The noise expressions correctly predict the dark
current noise behavior. The estimated gain and NEP from the shot
noise are in good agreement with the previous measurements.
(5)
ACKNOWLEDGMENT
p
(6)
where d is the thickness of a single emitter layer. The thermal
generation rate G (per unit volume) of a single emitter layer provides
carriers for internal photoemission and then have a probability (1 0 p)
The authors wish to express their gratitude to W. J. Schaff,
Cornell University, Ithaca, NY, and H. C. Liu, National Research
Council (NRC), Canada, for sample growth and device processing.
The authors also wish to acknowledge S. G. Matsik for his technical
help.
814
IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 46, NO. 4, APRIL 1999
REFERENCES
[1] A. G. U. Perera, “Physics and novel device applications in semiconductor homojunctions,” in Physics of Thin Films, vol. 21, M. H. Francombe
and J. L. Vossen, Eds. New York: Academic, 1995, pp. 1–75.
[2] E. E. Haller, “Advanced far-infrared detectors,” Infrared Phys. Technol.,
vol. 35, p. 127, 1994.
[3] D. M. Watson, M. T. Guptill, J. E. Huffman, T. N. Krabach, S. N.
Raines, and S. Satyapal, “Germanium blocked-impurity-band detector
arrays: Unpassivated devices with bulk substrates,” J. Appl. Phys., vol.
74, p. 4199, 1993.
[4] W. Z. Shen, A. G. U. Perera, H. C. Liu, M. Buchanan, and W.
J. Schaff, “Bias effects in high-performance GaAs homojunction farinfrared detectors,” Appl. Phys. Lett., vol. 71, p. 2677, 1997.
[5] W. Z. Shen, A. G. U. Perera, M. H. Francombe, H. C. Liu, M.
Buchanan, and W. J. Schaff, “Effect of emitter layer concentration on
the performance of GaAs p+ -i homojunction far-infrared detectors: A
comparison of theory and experiment,” IEEE Trans. Electron Devices,
vol. 45, p. 1671, Aug. 1998.
[6] B. F. Levine, “Quantum well infrared photodetectors,” J. Appl. Phys.,
vol. 74, p. R1, 1993.
[7] B. I. Shklovskii, “Theory of l=f noise for hopping conduction,” Solid
State Commun., vol. 33, p. 273, 1980.
[8] M. Y. Luo, G. Bosman, A. Van der Ziel, and L. L. Hench, “Theory and
experiments of l=f noise in Schottky-barrier diodes operating in the
thermionic-emission mode,” IEEE Trans. Electron Devices, vol. ED-35,
p. 1351, Aug. 1988.
[9] O. Jantsch, “Flicker (l=f ) noise generated by a randow walk of electrons
in interfaces,”IEEE Trans. Electron Devices, vol. ED-34, p. 1100, May
1987.
[10] R. C. Newman, “Local vibrational mode spectroscopy of defects in III/V
compounds,” in Semiconductors and Semimetals, vol. 38, E. R. Weber,
Ed. New York: Academic, 1993, pp. 117–187.
[11] L. Vadasz and A. S. Grove, IEEE Trans. Electron Devices, vol. ED-13,
p. 863, May 1966.
[12] J. Qiu, Q. D. Qian, R. L. Gunshor, M. Kobayashi, D. R. Menke,
D. Li, and N. Otsuka, “Influence of GaAs surface stoichiometry on
the interface state density of as-grown epitaxial ZnSe/epitaxial GaAs
heterostructures,” Appl. Phys. Lett., vol. 56, p. 1272, 1990.
[13] A. G. U. Perera, V. G. Silvestrov, S. G. Matsik, H. C. Liu, M. Buchanan,
Z. R. Wasilewski, and M. Ershov, “Nonuniform vertical transport and
relaxation in quantum well infrared detectors,” J. Appl. Phys., vol. 83,
p. 991, 1998.
[14] W. Z. Shen and A. G. U. Perera, “Photoconductive generation mechanism and gain in internal photoemission detectors,” J. Appl. Phys., vol.
83, p. 3923, 1998.
[15] D. Wang, G. Bosman, and S. S. Li, “On the dark current noise of
quantum well infrared photodetectors,” Appl. Phys. Lett., vol. 65, p.
183, 1994.
A Study of Rapid Photothermal Annealing on the
Electrical Properties and Reliability of Tantalum Pentoxide
Y. Chen, R. Singh, K. Rajan, D. J. Dumin,
S. DeBoer, and R. P. S. Thakur
Abstract—Rapid photothermal annealing is based on the use of vacuum
ultraviolet (VUV) photons as the source of optical energy and tungsten
halogen lamps as the source of optical and thermal energy. Tantalum
pentoxide (Ta2 O5 ) thin films deposited by thermal metalorganic chemical
vapor deposition (MOCVD) have been annealed by RPP and conventional
rapid thermal annealing (RTP). As compared to samples annealed by
RTP, lower leakage current and lower trap densities were observed in
the samples annealed by RPP.
Index Terms—Dielectric material, leakage current, rapid thermal annealing, reliability.
I. INTRODUCTION
For the future development of Si integrated circuits (IC’s), there is
a need for the replacement of silicon dioxide, silicon oxynitride, and
silicon nitride films by high dielectric constant (K) materials. From
device performance and reliability point of view, high K materials
need to have low leakage currents. From a process integration
and defect reduction point of view, low processing temperatures,
thermal, and residual stress constitutes some important requirements
[1]. Ta2 O5 films deposited by chemical vapor deposition (CVD)
techniques (bulk dielectric constant of around 25) have the potential
to meet the near future needs of silicon IC’s. However, relatively high
leakage currents [2], [3], widespread in the breakdown distribution
in tantalum pentoxide, and the lack of thorough understanding of the
reliability properties, hinders the use of Ta2 O5 film in manufacturing
Si IC’s. The degradation of Ta2 O5 films due to trap generation under
voltage stress is one of the main concerns for the scaled devices.
Rapid photothermal processing (RPP) [4] provides materials with
lower defects, better conformity, and lower thermal and residual
stresses in shorter processing time than other thermal processes.
In this paper, Ta2 O5 films deposited by thermal MOCVD have
been annealed by RPP and rapid thermal processing. For both
films, electrical characteristics like current-time (I 0 t), breakdown
distribution and trap density were measured and correlated to the
structural properties and thermal stress of the films.
II. EXPERIMENTAL
The metal-insulator-metal (MIM) structure was fabricated on 200mm n-type silicon substrates. The 120-nm bottom poly-Si electrode
was deposited by low-pressure chemical vapor deposition (LPCVD)
at 625 C followed by PH3 doping at 860 C. Prior to Ta2 O5
deposition, all samples were subjected to rapid thermal nitridation
Manuscript received April 3, 1998; revised December 1, 1998. The review
of this brief was arranged by Editor K. Shenai.
Y. Chen was with the Center for Silicon Nanoelectronics, Department
of Electrical and Computer Engineering, Clemson University, Clemson, SC
29634-5910 USA. He is now with Bell Labs, Lucent Technologies, Orlando,
FL 32819 USA.
R. Singh, K. Rajan, and D. J. Dumin are with the Center for Silicon Nanoelectronics, Department of Electrical and Computer Engineering, Clemson
University, Clemson, SC 29634-5910 USA.
S. DeBoer is with the Micron Technology Inc., Boise, ID 83706 USA.
R. P. S Thakur is with AG Associates, San Jose, CA 95134 USA.
Publisher Item Identifier S 0018-9383(99)02402-8.
0018–9383/99$10.00  1999 IEEE