I
DEFECT CHARACTERIZATION OF GAAS/ALGAAS MATERIALS AND
SILICONONINSULATOR DEVICES BY SEPARATIONBYIMPLANTOFOXYGEN
By
CHUNG GYUNE CHOI
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1988
TO MY FAMILY
FOR THEIR PATIENT SUPPORT
ACKNOWLEDGMENTS
I wish to express my great gratitude to Professor Sheng S. Li for his guidance and
encouragement throughout the research and preparation of this dissertation. I would also
like to thank professors D. E. Burk, Arnost Neugroschel, Gijs Bosman, and Murab Rao for
their participation on my supervisory committee.
I am grateful to Dr. Bob Loo in Hughes Research Laboratories for his fabrication of
AlGaAs devices. Thanks are extended to my friends and colleagues, Dr. TaeWon Jung,
Dr. Kerwin Teng, Ju Sung Park, Dr. JongSik Park, Dr. SangI Lee, Doo Hwan Lee,
Dr. JaeHoon Kim, Hong Shin Chen, Jin Young Choi, Hang Geun Jeong, Young Jun Yu,
Sang Sun Lee, Soon Young Huh, and YoungSuk Kim for their helpful discussions and
encouragements.
I am greatly indebted to my parents, my lovely wife and son, and my younger brothers
for their patience and endless love throughout this work.
TABLE OF CONTENTS
ACKNOWLEDGMENTS .............................................. iii
ABSTRACT ............................................................. vi
CHAPTER
ONE INTRODUCTION .............................................. 1
TWO ELECTRICAL CHARACTERIZATION METHODS ............. 4
2.1. CurrentVoltage Characteristics ............................ 4
2.2. CapacitanceVoltage Measurement ......................... 5
2.3. Thermally Stimulated Capacitance Method ................ 5
2.4. DeepLevel Transient Spectroscopy ........................ 6
2.5. Forwardbias Capacitance Spectroscopy .................... 8
2.5.1. Forwardbiased Schottky Diode ...................... 8
2.5.2. Dominant Exchanges with Semiconductor Conduction
Band ................................................ 9
2.5.3. Exchanges with Conduction and Valence Bands ....... 10
2.5.4. Exchanges with Conduction Bands of Semiconductor
and Metal .............................................. 12
THREE A GENERALIZED THEORY FOR DETERMINING THE
FIELDENHANCED THERMAL EMISSION RATE
BY REVERSE PULSED DEEP LEVEL TRANSIENT
SPECTROSCOPY ............................................. 13
3.1. Introduction ................... ........ .... ..... .. .. .......... 13
3.2. Development of A Generalized Theory ..................... 14
3.3. Determination of Temperaturedependent Capture Cross
Section ......................... ................. .............. 18
3.4. Summary ....... ................................... 19
FOUR AN ACCURATE DETERMINATION OF DX CENTER DENSITY
AND FREE CARRIER DENSITY IN ALxGAixAS ........... 22
4.1. Introduction ....................... ....................... 22
4.2. Theory ........ ...................................... 23
4.3. Summary .................. ................................. 30
FIVE PROTON AND ELECTRON IRRADIATION INDUCED DEEP
LEVEL DEFECTS IN ALGAAS ............................... 32
5.1. Introduction ................. ............................. 32
5.2. Theorectical Review of Defect Creation by Partical
Irradiation .......................... ............ .......... 33
5.2.1. Dynamics of Collision ............................... 33
5.2.2. Differential Scattering Cross Section ................ 36
5.2.3. Primary Displacements ............................. 36
5.2.4. Secondary Displacements ........................... 37
5.2.5. Formation of Amorphous Layer by Irradiation ......... 38
5.2.6. Range of Particle .................................... 39
5.3. Results and Discussion ................................... 40
SIX GROWNIN DEFECTS IN TEDOPED ALxGA_xAS GROWN
BY LIQUIDPHASEEPITAXY VS GROWTH TEMPERATURE
....................... .. .. ..... .................... 50
6.1. Introduction ............................................. 50
6.2. Experimental ............................................. 51
6.3. Results ............................................. 51
6.4. Conclusions .................................... ........ 56
SEVEN DEFECT CHARACTERIZATION AND ELECTRICAL
PROPERTIES
OF SIMOX BASED SOI DEVICES .......................... 59
7.1. Introduction .... ....................................... 59
7.2. Fabrication Process ........................................ 60
7.3. Oxygen Related Defects in Simox Based SOI Devices ....... 60
7.4. Device Characteristics versus Process Parameters ........... 74
7.4.1. Characterization Methods of SIMOX SOI Device ...... 74
7.4.2. Epilayer Thickness .................................. 75
7.4.3. Annealing Time .................................... 75
7.4.4. Lateral Isolation .................................... 76
7.4.5. Ion Dose and Annealing Temperature ............... 76
7.4.6. Defects and Electrical Properties of Pwell ............ 77
7.4.7. Nepilayer/pwell Interface .......................... 77
7.5. Summary and Conclusions ................................. 77
EIGHT DEVELOPMENT AND DEMONSTRATION OF AN ACCURATE
FORWARDBIAS CAPACITANCE SPECTROSCOPY .......... 83
8.1. Introduction ................ .. .. ........................ . 83
8.2. Accurate Forwardbias Capacitance Spectroscopy ........... 84
8.3. Results and Discussion ................................... 88
NINE CONCLUSIONS AND RECOMMENDATIONS ................. 91
APPENDIX ........................................................ .... 94
REFERENCES ......................................................... 95
BIOGRAPHICAL SKETCH ............................................ 99
Abstract of Dissertation Presented to the Graduate School of
the University of Florida in Partial Fulfillment of the Requirements
for the Degree of Doctor of Philosophy
DEFECT CHARACTERIZATION OF GAAS/ALGAAS MATERIALS AND
SILICONONINSULATOR DEVICES BY SEPARATIONBYIMPLANTOFOXYGEN
By
Chung Gyune Choi
April 1988
Chairman: Sheng S. Li
Major Department: Electrical Engineering
This dissertation describes the defect characterization of GaAs/AlGaAs epitaxial lay
ers and SiliconOnInsulator (SOI) devices by Separation byImplantofOxygen technique.
A novel experimental theory was developed for this study. A generalized theory for de
termining the fieldenhanced thermal emission rates from the deep level defects using the
Reversepulsed DeepLevel Transient Spectroscopy (RDLTS) has a potential as a powerful
characterization method for the decision on the potentialwell types of deeplevel defect
states. This method enables us to measure the fieldenhanced thermal emission rates at
very high electricfields. The previous methods cannot give those experimental data.
An experimental theory for determining the temperaturedependent capture cross sec
tion was developed as the byproduct of a generalized theory. This method is simpler and
easier than the previous experimental methods.
An accurate determination method for the DX center and shallow center densities was
developed. This method utilizes the thermally stimulated capacitance (TSCAP) technique.
Under the carrier freezeout condition the DeepLevel Transient Spectroscopy measurement
cannot give the exact DX center density.
To make progress in the area of defect characterization, a generalized theory for de
termining the fieldenhanced thermal emission rates must be applied. This method makes
use of the fast part of capturing transient and hence the DLTS system must have the fast
circuitry. The computerized DLTS system with the fast circuit option is commercially avail
able. If this fast DLTS system utilizing a generalized theory is applied, it gives a great deal
of flexibility in defect characterization.
An accurate forwardbias capacitance spectroscopy was developed for the study of the
Schottky interface state. This method is very accurate and efficient compared the previous
methods. An accurate forwardbias capacitance spectroscopy is implemented and is fully
demonstrated by measuring the real Schottky diode of small area. This method makes a
great breakthrough in the area of the metalsemiconductor interface state study.
CHAPTER ONE
INTRODUCTION
The GaAs and its related compound semiconductor materials have shown promise in
the applications of faster and denser integrated circuit designs. The epitaxial growth tech
nologies such as molecular beam epitaxy (MBE), metalorganic vapor epitaxy (MOCVD)
and liquid phase epitaxy (LPE) have become the major fabrication methods for the GaAs
and its related compound semiconductor devices. The device and materials characterization
must be carried out for quality control.
Defect plays the dominant role in determining the overall device performances. The
grownin defects should be controlled below the tolerable levels. The processrelated defects
should be traced back to their origins in order to develop a new process step of evading
those defect levels. Nowadays the fabricated devices are required to have the radiationhard
properties because the devices might operate in the radiation environment such as in the
communication satellite. The irradiationinduced failure mechanisms in the GaAs integrated
circuits should be investigated as part of the process development for the radiationhard
devices. Intensive study on the proton or electron induced defects in GaAs has been
carried out. However, very few experimental data concerning the radiationinduced defects
in the AlGaAs layers are available.
Many theoretical papers about the potentialwell types have been published because
the decision on the potentialwell type of a defect helps us to find out the defect origin. In
order to determine the potentialwell type of a defect one must measure the fieldenhanced
thermal emission rates versus electric field. The electric field range should be from very low
to very high. However, the published data were limited to rather low electric field. This
was caused by the absence of adequate experimental theories and techniques. A generalized
theory for determining the fieldenhanced thermal emission rate has been developed by
the author. Using this theory, the data of very high electric field is, for the first time,
successfully obtained. This generalized theory enables us to determine the potenitalwell
type of defect.
The determination method for temperaturedependent capture cross section is devel
oped as the byproduct of the generalized theory. The activation energy of capture cross
section can be determined in a much easier and simpler way than the previous methods.
The measured capture cross section can be used for determining the fieldenhanced thermal
emission rate selfconsistently.
An accurate determination method for the DX center and shallow donor densities,
especially in AlxGaixAs (x > 0.3), is developed. In addition, an accurate determination of
free carrier concentration is performed by considering the Hall electron concentration and
the ratio of Hall to drift mobility. The previous method usually underestimates the shallow
donor density in the AlxGaixAs of Aluminum composition over 30 %.
The epitaxial growth of GaAs on silicon substrate has become the important research
subject. The excellent devices with GaAs/AlGaAs epitaxial structures grown on Si have
been fabricated [17]. The microwave performance of GaAs metalsemiconductor field
effect transistors on Si has been shown to be as good as MESFET's on GaAs substrate.
The GaAs integrated circuit on Si substrate is considered as a costeffective substitute for
GaAs bulk grown wafers, especially for digital or lowfrequency microwave applications.
Silicon is a much better thermal conductor than GaAs, and thus GaAs on Si can be applied
to the various applications such as power FETs. As mentioned earlier, one of the failure
mechanisms in GaAs digital integrated circuits in the radiant environment is the conversion
of semiinsulating substrate into semiconducting layer. In order to prevent this problem,
GaAs on SOI (silicon on insulator) substrate is proposed as the future device stucture.
The presence of the oxide layer nullifies the necessity of the semiinsulating layer and it is
possible to fabricate the less thick epitaxial layer because the buffer layer is probably not
necessary in this new device structure. It also improves the device switching speed because
the parasitic capacitance reduces. The SOI substrate structure can be easily fabricated by
the SIMOX (Separation by Implant of Oxygen) technique. The SIMOXbased SOI devices
were fabricated and the electrical characterization was performed by using the Deep Level
Transient Spectroscopy, currentvoltage, and capacitancevoltage methods. The process
related defect levels such as divacancy plus oxygen complex were observed in the epilayer
grown on the superficial layer which went through high temperature annealing steps.
An accurate forwardbias capacitance spectroscopy was developed for the study of the
Schottky interface state study. This method is very accurate and efficient compared to
the previous methods. They had the difficulty to measure the forwardbias capacitance
under the high current flow condition. An accurate forwardbias capacitance spectroscopy
is implemented by the author and if fully demonstrated by measuring the real Schottky
diode of small area. This method makes a great breakthrough in the area of the metal
semiconductor interface state study.
CHAPTER TWO
ELECTRICAL CHARACTERIZATION METHODS
2.1. CurrentVoltage Characteristics
Currentvoltage characteristics of pn junction under forwardbias condition is given
by
Jt = Jp(x.) + J(xp) (2.1)
= qn( D + )(ea/kT 1) (2.2)
NdLp NaLn
= Jo(e v/kT 1) (2.3)
where Jp(xn) is the minority carrier (hole) current in the nregion and Jn(xp) is the minority
carrier (electron) current in the pregion, respectively; xn and Xp are the depletion layer
edges in the nregion and in the pregion, respectively; Jo is the saturation current density
and Va is the applied forwardbias voltage; Dn(Dp) is the diffusion constant for electrons
(holes); and Ln (Lp) is the diffusion length for electrons (holes), and Na(Nd) is the shallow
acceptor (donor) density. If the recombination current component is considered in the
spacecharge region, the recombination current Jr can be expressed by
qxdni Va
Jr = exp( V (2.4)
2r. 2kT
where ro is the lifetime associated with the recombination of excess carriers in a region with
a density of Nt of recombination centers. The lifetime 7o is expressed by
To = N (2.5)
NtOvth
The total forward current can be approximated by the sum of Eqs. (2.1) and (2.2) for Va
> kT/q:
F qDpn qVa qxdni / Va
JF = q dexp() + exp( (2.6)
F7p Nd kT 2ro 2kT
This total forward current can be represented by the empirical form,
JF = Joexp(qVa) (2.7)
nkT
The factor n is equal to 2 if the recombination current dominates and n is equal to 1 if the
diffusion current dominates. When both current components are comparable, the ideality
factor n has a value between 1 and 2.
2.2. CapacitanceVoltage measurement
The depletion layer capacitance is defined as C = dQ/dV, where dQ is incremental
charge increase according to an incremental change of the applied voltage dV. For the
onesided abrupt junction or Schottky barrier diode, the capacitance is given by
dQ As\
C A (2.8)
dV W
Sb(Vb V 2kT)1/2 (2.9)
= 2 q
where c, is the dielectric constant of semiconductor; Vbi is the builtin potential and V is
the applied voltage where signs are for the reverse and forwardbias conditions, respec
tively. NB is the background doping concentration. One can deduce the background doping
concentration from Eq. (2.6) and using the relationship,
d(/C2) 2 dC10)
= ( (2.10)
dV C3 dV
so that,
N(xd) 2 (2.11)
2
N(X) = q,[d(1/C2)/dV]A2 (2
2.3. Thermally Stimulated Capacitance Method
The TSCAP measurement is carried out by first cooling the junction device to 77 K and
then zero biasing it to fill the majority carrier traps. The junction is again reversebiased
and then temperature is raised from 77 K to 400 K [8]. The thermal scan of capacitance
versus temperature plot is then taken by using an xy recorder. A capacitance step is
proportional to the trap density. The trap density can be deduced by the relationship
2AC
Nt = Nd (2.12)
Co
where Co is the depletion layer capacitance and AC is the capacitance change due to the
majority or minority carrier emission. Eq. (2.12) is valid only for the case when Nt is much
less than Nd. For the case of large trap density, a more exact expression should be used [9].
C2 A2qEOEs NT (2.13)
2 1 2(VD + VR)
This expression will be explained in detail in Chapter Four.
2.4. DeepLevel Transient Spectroscopy
The Deep Level Transient Spectroscopy (DLTS) is a high frequency (20 MHz) transient
capacitance technique which is first introduced by Lang in 1974 [10,11]. The DLTS scan
displays the spectra of deep level traps in the forbidden gap of a semiconductor as positive
or negative peaks depending on minority trap or majority trap. From this measurement it is
possible to measure the following parameters such as trap density, activation energy, defect
concentration profiling, electron and hole capture cross sections. Though the capacitance
transient were made use of in the conventional DLTS, the modified DLTS methods such
as current DLTS, optical DLTS, and RDLTS (Reversepulsed DLTS) have been developed
according to their specific needs subject to the test device structures and characteristics
[1216]. For simplicity, the theory of conventional DLTS will be explained briefly. The
transient capacitance can be expressed as
C(t) = Co[1 exp(t/r)]1/2 (2.14)
where Co = C(Vr) is the depletion layer capacitance and r is decay time constant. This
time constant will be shown to be the inverse of emission rate as follows. Eq. (2.9) reduces
to a simpler form by using binomial expansion if the defect density is less than one tenth
of background doping, Nd
1Nt
C(t) = Co[1 Texp(t/r)] (2.15)
2 Nd
Here, Eq. (2.10) can be rewritten as
Ntexp(t/r)= C(t)Nd (2.16)
Co
where AC(t) = Co C(t). From the DLTS measurement, AC(0) can be obtained. The
junction capacitance Co and the background doping concentration, Nd, can be deduced
from the CV measurement. Thus, the defect density Nt, can be calculated by using Eq.
(2.11). Eq. (2.11) can be rewritten as
C(t) = AC(0)exp(t/r) (2.17)
Then, the capacitance difference can be defined by using two different time points, tl and
t2
S(r) = C(0)[exp(ti/r) exp(t2/r)] (2.18)
The maximum or minimum value of S(r) is obtained by differentiating S(r) with respect to
r, i.e. dS(r)/dr = 0, which yields
(tl t2)
S= max ( (2.19)
ln(ti/t2)
This means that S(r) reaches to a maximum or minimum value at a given window rate
(i.e. rmax) and this peak point corresponds to a temperature point in the DLTS spectrum
which is a function of S(r) vs temperature. This temperature corresponds to an emission
rate which is the function of temperature. The emission rate is given by
en = Tma.x
=, Tma
(2.20)
The emission rate can be expressed by
Ec Et
en = (an < Vth > Nc/g)exp( kT ) (2.21)
where Et is the trap activation energy; g is the degeneracy factor; an is the electron capture
cross section which is temperature dependent as [17]
AEb
n = axp( (2.22)
where aoo is the capture cross section at very high temperature; AEb is the activation
energy of the capture cross section. The emission rate of Eq. (2.16) can be rewritten as
Ec (Et + AEb)
en = BT2exp (2.23)
kT
Arrhenius plot of ln(en/T2) versus 1/kT can be carried out from several DLTS spectra. The
slope of this plot is the apparent activation energy (i.e., Et + AEt).
2.5. Forwardbias Capacitance Spectroscopy
2.5.1. Forwardbiased Schottky diode
In a forwardbias Schottky diode, the charge of the interface states can be electrically
modulated. The modulation of charge state variation with frequency, bias, and temperature
appears as the capacitance and conductance change [18,19]. It is generally accepted that the
quasiFermi level of majority carriers is flat throughout the spacecharge region of a forward
biased Schottky junction. The position of the quasiFermi level at the interface varies with
bias [20]. On the other hand, in reverse bias, the quasi Fermi level of majority carriers
remains pinned on that of metal. In forwardbias condition, the relationship between the
quasiFermi level and that of metal at the interface is given by
EFNS EFM = qV (2.24)
and the free carrier density at the interface is expressed by
Ec EFNS
n, = Ncexp( kT (2.25)
kT(225
In a forwardbias condition, the conductance due to the thermionic current Gsc = qI/kT
is very high and entirely obliterates Gi, due to the interface charge variation [18,20]. On
the other hand, Cc, ( capacitance due to charge modulation on the edge of the spacecharge
area) varies only slowly with bias and Cis (due to the interface charge variation) exceeds
Cos to a large amount. If a small signal v = vo exp(iwt) with Vo << kT/q, is superimposed
on the direct bias voltage V, the admittance due to the state density Nis can be expressed
by [19]
Yis = Gis + iwCis (2.26)
where
Cis= q 2 N j(1 F)cnns (2.27)
1 + W272 kT
T1 = Cnn, + Cnnl + Cpps + Cppl + rT1 (2.28)
The cutoff frequency will be
1
fc = (2.29)
Now we can derive the important parameters such as the interface state density activation
energy and capture cross section using the above equations in low frequency domain and at
constant temperature.
2.5.2. Dominant exchanges with the semiconductor conduction band
(cnni >> Cpps, rT).
The occupancy F and carrier exchange time constant can be simplified as
F= n (2.30)
n, + ni
r 1 = Cnn, + cnni
(2.31)
If q4 qV, = Es(V = V,) holds, then the electron quasiFermi level EFNS merges with the
interface states E.. The interface capacitance will be expressed as
q2Ni.
Ciso = F(1 F) (2.32)
The maximum peak condition of dCio/dF = 0 gives F = 0.5. At F = 0.5, Ciso has a peak
value. The activation energy E, of interface state can be determined as shown in Fig. 2.1.
The cutoff frequency is given by
Scn(n + n) (2.33)
fe= (2.33)
27r
Therefore, the cutoff frequency is proportional to Cnni (if V < Vs) and is proportional to
Cnns (if V > Vs). If the interface state is a continuous band states, then the spectrum of
Ciso will be broad.
2.5.3. Exchanges with the Conduction and Valence Bands
In this case, the occupancy F and the carrier exchange time constant can be simplified
as
F cnns (2.34)
cnn, + Cpps
r1 = cnns + Cpps (2.35)
Likewise as in the first case, at F = 0.5 the spectrum of Ciso has a maximum value located
at the demarcation energy level Edn is defined by
cpNv
Ec Edn = cBp kln(cN) (2.36)
cnNc
The behaviour of a band state depends on its relative location with respect to the demar
cation level Ed, If E,2 > Es > Ed, the direct observation is possible. If Edn > Es2 > Esl,
the apparent spectrum is centered on Edn. If Es2 > Edn > Esl, the apparent spectrum is
truncated and is reduced to the range from Ed, to Es2.
Ciso
Ciso
Figure 2.1.
E (qVF)
Esl Es2
Interface capacitance spectrum with the position of the quasiFermi
level of majority carriers at the interface: (a) Discrete interface state
and (b) Band of interface state.
E (qVF)
2.5.4. Exchanges with the Conduction Bands of Semiconductor and Metal
Cnns
F = c(2.37)
Cnns + TT
r1 = cnn, + rf (2.38)
Ec Edn = kTln(cnNcrr) (2.39)
The peak of Ciso spectrum occurs at Edm. The actual activation energy of state can be
obtained from the slope of log fc vs E curves. The reemission of electrons to metal is
dominant if the Ciso peak moves with temperature, whereas the reemission of electrons
to the conduction band of the semiconductor is dominant if Ciso is independent of the
temperature.
CHAPTER THREE
A GENERALIZED THEORY FOR DETERMINING FIELDENHANCED
THERMAL EMISSION RATE BY REVERSE PULSED DEEP LEVEL
TRANSIENT SPECTROSCOPY
3.1. Introduction
A generalized theory for determining the fieldenhanced thermal emission rates and
carrier capture cross section of deep level defects at very high field and for large trap
density by the reverse pulsed deep level transient spectroscopy technique is developed in
this chapter. Using this new theory the fieldenhanced emission rates for the DX center in
a liquidphase epitaxy grown Sndoped Alo.2Gao.sAs were determined for field strength of
up to 7 x 105 V/cm.
In this study we report the derivation of a generalized theory for determining the field
enhanced thermal emission rates at very high field and for arbitrary ratios of NT/ND using
the RDLTS technique. It is wellknown that determination of potential well for a deep level
defect is useful for identifying its physical origin. The theories for determine the types of
potential well for the deep level traps in a semiconductor from the fieldenhanced emission
rate have been published in the literature [2125]. However, the experimental data on the
fieldenhanced emission rates reported so far have been limited to rather low field due to the
lack of adequate theory for describing the fieldenhanced rates at very high field and large
trap density. As a result, the fieldenhanced emission rate data for field strength greater
than 2 x 105 V/cm have not been reported in the literature.
Recently Li and Wang developed the RDLTS technique assuming that the transient
capacitance signal comes mainly from the narrow junction spacecharge region at the in
tersection of ET = EF, and its amplitude is proportional to the density of deeplevel trap
[26]. However, as will be explained later in the text, the RDLTS signal is not a simple
exponential transient; instead it composes the fast and slow capacitance transients. In fact,
the capturing process occurs in a broader region from the intersection of ET = EF to the
depletionlayer edge. In addition, in the case where the defect density is comparable to the
shallow dopant density, the relationship, NT/ND = 2 AC/Co, is no longer valid. Therefore,
a rigorous theorectical model is needed in order to obtain the data of fieldenhanced emis
sion rates at high fields and arbitrary trap density from the RDLTS technique. A general
expression for the charge state of a deeplevel trap under the transient condition will be
derived next. From this equation the fieldenhanced emission rate at very high field and for
arbitrary trap density can be determined.
3.2. Development of A Generalized Theory
For the convenience of discussion, a Schottky barrier diode will be used for deriving
the RDLTS theory. Figure 3.1 shows the energy band diagram of a Schottky diode in
thermal equilibrium and under reversebias conditions in the RDLTS measurement. W1 is
the depletion width under steadystate condition, and W1x is the distance from x = 0 to
the point where EF coincides with ET; Wi(t) and Wlx(t) are the corresponding parameters
under transient condition. If the defect density is comparable to the shallow donor density,
then the initial depletion width W1(0) in the transient state will be drastically reduced from
W1 due to the increase of ionized traps during the reversebiased pulse. W2 and W2x are
the depletion width and the distance from x = 0 to the point where ET = EF, respectively,
when a reversebiased pulse of V2 is applied. During this reversebiased pulse of width tp,
the occupied traps emit electrons to the conduction band and unoccupied traps captrue
electrons from the conduction band. Following the reversebiased pulse of duration tp, the
applied voltage is returned to the quiescent voltage of V1 and the transient state follows.
The charge state of the trap in the transient state can be expressed by [27];
= r1N + en(NT N+) (3.1)
I I
~0
o ~0 0
L w
I I
II I
2
0 W2x I
S* *0 *
(b) i o
1 0 E
I Ev
*i O  ,,
0I o0 E c0
E F
ET
S 14dW W1 W( Ec
 EF
I Ev
wP( I
1w 1x M
I I
I_
X=O
Figure 3.1.
Energy band diagram of a Schottky barrier diode under different
reversebiased pulse conditions: (a) at reversebias voltage Vi under
steadystate conditions, (b) at V2 during the reversebiased pulse tp,
and (c) at V1 following the reversebiased pulse and in the transient
state.
(a)
E,
EF
ET
Ev
t t
t=O
or
dN+
d + (7r + e^)N+ = enNT, (3.2)
where
r, = anVthn(x), (3.3)
n(x) = noexp[Wl(t) x]2/2LD, (3.4)
en = UnVthnoexp[(ET EF)/kT], (3.5)
and LD is the extrinsic Debye length.
In Eq. (3.1), since both rT' and W1(t) vary with time, NT+(x,t) cannot be expressed
by a simple exponential function. Instead, r'(x,t) is a fast varying function of position and
time when the defect density is comparable to the shallow donor density and the depletion
layer edge changes rapidly. Thus, by multiplying Eq. (3.2) by I(x,t), one obtains
d Ixt)] = I(x,t)enNT, (3.6)
dt
where
I(x,t) = exp (r1 + en)dr. (3.7)
Solving Eq. (3.6) yields
N+(, I(x,t) 1 I(x, r)enNTdr + I(,t (3.8)
N I(x, t) oxt) (3.8)
Eq. (3.8) is an exact expression for the ionized deeplevel defect density. Note that the value
of C1 in Eq. (3.8) can be determined from the initial conditions as follows. In the region
[W1x(O) < W1x], the defects remain totally ionized since ET is greater than EF during the
reversebiased pulse period tp.
N+(x,O) = NT for W1x(O) < x < Wlx.
(3.9)
In the region (Wx < x < Wx,), the deeplevel defects experience the fieldenhanced thermal
emission during the reversebiased pulse of pulse width tp,
N+(x,0) = NT 1 exp(entp)] for W1x < x < W1 (3.10)
From Eqs. (3.8), (3.9) and (3.10) one can obtain the expressions
N+(x,t) _x) 1 ft d NT (3.11)
I(x,t) (xr)enN I(x,t)
for W1x(0) < x < Wx, and
N(x, t I(x,r)enNTdr + NT (3.12)
NI(x,t) oI(x,t)
for W1, < x < W1.
In Eqs. (3.11) and (3.12), the steadystate condition (i.e., t = oo) is also satisfied, and
N (x,)= en NT. (3.13)
T (r'1 + en)
Eq. (3.13) can also be derived by setting dN+/dt = 0 in Eq.(3.1). Thus, it is clear that
Eqs. (3.11) and (3.12) are the exact expressions for the density of the empty state defect
in the transient as well as steady state conditions.
To derive a generalized expression for determining the fieldenhanced emission rates by
the RDLTS measruement at very high field and for arbitrary trap density, one can start
from the Poisson equation by using Eqs. (3.11) and (3.12). This is given by
C W1. (0)
q(V Vb) (ND + NT)xdx
/ix 1 t t NT
+ (ND + I(x, r)enNTdT + )xdx
w1. (0) I(x, t) o I(x, t)
w'ix(t) 1 t NT[1 exp(entp)]
+ (ND + x t I(x, )eNTd + xt) )xdx (3.14)
Wix I(x,t) J I(x,t)
In Eq. (3.14), en and En are different in that en denotes the thermal emission rate at low
field and in the transient state conditions, and En is the fieldenhanced thermal emission rate
during the reversebiased pulse. Since the region [W1 < x < W2] can be neutralized within
the dielectric relaxation time, tD, this region is not considered in Eq. (3.14) assuming that
tD/Tc = EanVn/q/Pn is much less than unity [28].
Equation (3.14) can be differentiated with respect to time by using the Reibniz rule,
Op u2 Of du2 dul
Ot U dx + f(u2, t) f(ux, t) (3.15)
where
p(t) = f(x, t)dx, (3.16)
and the result is given by
a 
( (V + Vbi)) = 0
Ot q
wx(t) (r'1+ en) t NT(T + en)
7c I(x, r)enNT d + enNT )xdx
JW.(o) I(x, t) o II(x, t)
S1 fo NT[1 exp(entp)] dW1
+ (ND + I(xtf I(x, r)enNTdr + (t) )W(t)
I(xt) I(xt) dt
+ W NT(r 1 + en) exp(entp)xdx. (3.17)
JWix I(x, t)
For t = 0, Eq. (3.17) reduces to
fr (0) NTrclxdx (ND + NT)WI(0)(W )lt=o
exp(entp)= WI.() dt (3.18)
ww f (0) NT(r,1 + en) x dx NTWI(0)(dIt )lt=o
In Eq. (3.18), W1(0), W1x(0), W1(0), and dWi/dt can be measured and f rlxdx can
be solved analytically as shown in the Appendix. Therefore, by using the relationships,
Wi(0) = cA/Ci(0) and Wi(t)/dt = (cA)2/C1 3dC1/dt, one can obtain the plot of en
versus electric field. The corresponding electric field with en can be determined by using
the depletion approximation with the reversebiased voltage V2.
3.3. Determination of Temperaturedependent Capture Cross Section
If entp approaches infinity, then exp(entp) goes to zero. This can be achieved under
a very large reversebiased pulse condition for a long duration pulse. This would allow the
traps to be totally ionized, and hence the transient capacitance signal can reach saturation.
From Eq. (3.18) and using the relation that exp(entp) = 0, the following expression is
obtained:
/Wi(0) ND dW1
,(o) r dx = ( + 1)WI(0)t=o, (3.19)
JWix(o) NT dt
where
1 [Wi x]2
T, avthnoexp ]2 (3.20)
and
1 w wl(o) [(W(t) x]
a [ (0) exp( t) )xdx] (3.21)
vthno JW1ix() 2L
ND dW1
X (T + 1)W1(0) d t=o.
NT dt
Here one can measure the capture cross section at a fixed temperature by using Eq. (3.21).
From measurements of the capture cross sections at various temperatures one can plot
a versus 1/kT. In this plot, both aoo and EB can be determined by using the following
expression for the temperaturedependent capture cross section [29].
a = aoexp( ). (3.22)
If this capture cross section is applied to Eq. (3.18), then the fieldenhanced thermal
emission rate at any temperature can be determined selfconsistently.
3.4. Summary
We have performed the RDLTS measurement on the LPEgrown Sndoped Alo.2Gao.sAs
sample, and the fieldenhanced emission rates for the DX center have been determined
from the theory presented in this text, as shown in Fig. 3.2. In the experiment the fast
capacitance transient was observed in the 1 p, second time frame [(r,1 = (avthno)1 = 3.42
x 106 s]. The observed value of capture cross section, a is in good agreement with the
value reported by Bhattacharya et. al [30]. From the CV and DLTS measurements ND =
2.92 x 1017 cm3 and NT = 3.10 X 1016 cm3 were also determined for this sample.
20
A generalized theory for determining the fieldenhanced thermal emission rate at very
high electric field and for arbitrary NT/ND ratio by the RDLTS technique has been derived
in this chapter. The theorectical expressions for the fieldenhanced emission rates were
derived by using the fast capacitance transient and its slope in the RDLTS measurement. In
addition, it is shown that the capture cross section can be determined by using the saturated
capacitance transient which is obtained by applying a high reversebiased saturation pulse.
The presented theory is applied to determine the field enhanced thermal emission rates for
the DX center in the LPEgrown Alo.2Gao.sAs sample.
I I I I I I 
Sn doped
T=100 K
AlO.2GaO.8As
, *'
J.
I
I
/
0
I
I I I I I I I I
1 2 3 4 5 6 7 8 9
ELECTRIC FIELD
(X105
V/cm)
The fieldenhanced thermal emission rates versus electric field for
the Ec 0.20 eV (i.e., DX center) in the LPEgrown Alo.2Gao.sAs
as determined by the RDLTS measurement at 100 K.
I
C,)
102
Figure 3.2.
103
CHAPTER FOUR
AN ACCURATE DETERMINATION OF DX CENTER
AND FREE CARRIER DENSITIES IN ALxGAixAS
4.1. Introduction
Accurate determination of the densities of DX center and free carriers in AlxGalxAs
(for x > 0.3) grown by liquidphase epitaxy (LPE) has been made by using the deep
level transient spectroscopy (DLTS), thermallystimulated capacitance (TSCAP), constant
temperature capacitancevoltage (CV) and Halleffect measurements. An anomalously high
density of DX center is determined by the TSCAP and lowtemperature CV techniques
for AlxGalxAs under the condition that majority of carrier freezeout occurs at 77 K. The
apparent carrier density determined by the CV method at room temperature is found to
be equal to the sum of the DX center density and the shallow donor density.
It is well known that the dominant deepelectron traps known as the DX center plays
a major role in controlling the electrical characteristics of AlxGalxAs alloy system for x
> 0.2. Caswell et al. [31] showed recently that the ionized donors were not observable
below 150 K in the MBE grown Sidoped Alo.35Gao.65As. A significant decrease of the Hall
mobility and free carrier concentration as well as a marked carrier freezeout are observed
if the Al content is increased beyond 0.25 [32,33]. Since the overall concentration of deep
electron traps is greatly enhanced near the directtoindirect bandgap crossover point [34],
it is important to study the effect of deep electron traps of ntype AlxGalxAs with x range
between 0.30 and 0.50. Mizuta et al. [35] and Bhattacharya et al. [36] showed that the DX
center is a substitutional deepdonor state in which its activation energy is associated with
the L conduction band minima; Kunzel et al. [37] pointed out that for x = 0.35 the DX
center governs the electrical properties of AlxGaixAs. Nevertheless, for x > 0.3, the density
of DX center may exceed the background density and the Hall electron concentration should
include the contribution of three different valleys (i.e., F, L, and X bands). Therefore, an
accurate determination of the DX center density and the free carrier concentration becomes
extremely difficult for x greater than 0.30.
4.2. Theory
In this dissertation, an accurate method of determining the densities of DX centers
in LPEgrown Tedoped and Sndoped AlxGalxAs (0.2 < x < 0.4) using the combined
DLTS, TSCAP and constant temperature CV method is presented. In addition, an accurate
determination of the free carrier concentration is performed by considering the Hall electron
concentration and the Hall to drift mobility ratio due the three conduction valleys. It
should be noted that the density of DX center increases dramatically as the aluminum
mole fraction ratio x is greater than 0.3. Our lowtemperature CV measurements revealed
that a complete carrier freezeout occurred at 77 K for the Sndoped Al0.33Gao.67As and
Al0.4Gao.6As samples as shown in Fig. 4.1 and Fig. 4.2. This result does not necessarily
mean that the shallow donor centers are not available at 77 K. Still, a good portion of
the shallow donor centers can exist. This can be explained in terms of the deepening of
the shallow donor activation energy. This will be shown later in this paper. As for the
roomtemperature CV measurement, the junction capacitance is contributed both from
the shallow and the deepdonor centers because the electron capture and emission at the
DX center can follow the 1 MHz small signal at room temperature. However, in the neutral
region the deep donors are mostly occupied since the Fermilevel is located close to but
slightly above the DX center level.
In the case that the concentration of DX center is very large, the simple relation NT =
2[AC(0)/C]ND used in the conventional DLTS analysis is no longer valid. Thus, for high
Al composition, it is impossible to determine the exact density of DX center by the DLTS
measurement due to the carrier freezeout at low temperature. To provide an accurate
30 rPuJJ.1 LL I10"33 T0"67.Mb
U 20
20
I '^300 K
10 160 K
10
120 K
77 K
0 2 4 6 8 10
REVERSE VOLTAGE (V)
Figure 4.1. The constanttemperature capacitancevoltage measurements of
Alo.33Gao.67As and the complete freezeout at 77 K.
2 4 6 8
REVERSE VOLTAGE (V)
The constanttemperature capacitancevoltage measurements of
Alo.4Gao.6As and the complete freezeout at 77 K.
Figure 4.2.
determination of the density of DX center, the TSCAP method can be used and a new
equation has been derived, which is given by [38,39]
c2 c2 A2qCo's
A2qOES NT (4.1)
Scl 2(VD + VR)
where C1 and C2 are the values of capacitance taken from the capacitance step observed in
the TSCAP scan as shown in Fig. 4.3. The TSCAP curve clearly shows two capacitance
steps for the Alo.33Gao.67As sample, and the locations of capacitance steps coincide well
with those DLTS peaks for the DX centers shown in Fig. 4.4.
The free electron density for the Tedoped Alo.38Ga0.62As determined by the CV
method is 1.54 x 1017 cm3 which is much higher than the electron density nh of 1.5
x 1016 cm3 determined by the Hall effect measurement at room temperature. The reasons
for this discrepancy can be attributed to : (1) The DX center, whose density is 1.41 x 1017
cm3 as determined from our new method, plays a major role, and (2) the Hall electron den
sity should be corrected by the Hall to drift mobility ratio including muliticonduction band
structure. The Hall electron density and the total free electron density can be expressed
respectively by [40,41]
nr[1 + + X n L1]2
nh = n[1 n+nr rur )2 (4.2)
h JX+ PCX(J1)2 4+ n.L L/)2J
nr r 'r nr nr r
nx nL Ph
nt = nr + nx + nL = nr[1 + + ] = nh (4.3)
nr nr Pd
where Pr, px and PL are the mobilities in r, X, L conduction band minima, respectively.
Assuming that the Boltzman statistics is valid for the present study, the following relation
ships also hold
nr = Nce kT = (2 eT )3 (4.4)
m* = 0.067 + 0.083x (4.5)
nL,x (mxErx ,r
n( X)2 exp( (4.6)
nr mr kT
150
200
250
300
TEMPERATURE (K)
The thermallystimulated capacitance of the Sndoped A10.33 Gao.67As
with two capcitance steps which means two DX centers.
0
H
0
0
77 100
Figure 4.3.
150
200
250
TEMPERATURE (K)
The DLTS spectra of the Sndoped Alo.33Gao.67As with two
DX centers.
77 100
Figure 4.4.
300
nr + nx + nL + X = NSD + NDX NA (4.7)
1 + e kT ,
where EFL and Erx are the FX and FL intervalley separation. mdX,L is the density of state
effective mass in the X and L bands. The computer simulation for the ratio of Hall to drift
mobility can be executed by taking into account of all the possible scattering mechanisms in
this alloy system using the published physical parameters [40,41]. The calculated value of nt
at 300 K in the Tedoped Alo.3sGao.62As is 2.2 x 1016 cm3 based on Eq.(4.3). In Eq.(4.7),
the righthand side term can be determined from the roomtemperature CV measurement.
Also, the DX center density NDX can be accurately determined from Eq.(4.1). Thus, the
shallow donor density can be determined from the difference between the roomtemperature
CV measurement and the DX center density determined by Eq.(4.1). The measured shallow
donor density is 1.3 x 1016 cm3. This means that both the shallow donor centers and the
DX centers comparably supply the free electrons to the three conduction bands at room
temperature. Thus, the widespread understanding that the free electron densities at the
multiconduction bands are supplied by the DX centers, is probably incorrect. The location
of the Fermilevel can be determined from Eq. (4.7). The Fermilevel is located close to
but slightly over the DX center level at room temperature. The results are summarized in
Table 4.1.
In addition, the validity of using CV method for determining the shallow donor density
and the DX center density must be reviewed. This is discussed as follows. The CV data
at 300 K gives the sum of the shallow and DX center densities, and the 77 K CV data
gives the shallow donor density. The difference between them is equal to the DX center
density. This method is based on the assumption that the Fermilevel is located below the
shallow donor level at 77 K and the shallow donor centers are completely ionized at 77 K.
However, this assumption is incorrect in the AlxGalxAs for x > 0.3 in which the shallow
donor energy becomes deeper based on the hydrogenic model. Using Eqs. (4.4) and (4.5),
the Fermilevel location can be determined at 77 K. Since the shallow donors are associated
with the r band and other deep donors are totally frozen at 77 K, the relationship nt =
nr holds. If it is assumed that the shallow donor density is in the range of 1014 to 1016
cm3, then the corresponding location of the Fermilevel, which is calculated by Eq.(4.4),
will be in the range of 0.0107 eV < Ec EF < 0.0413 eV. Based on the published data,
the activation energy of the shallow donor center reaches 17 meV at x = 0.31 [42]. This
means that at 77 K the Fermilevel is located above or close to the shallow donor energy
and the shallow donor centers cannot be completely ionized in the AlxGai_xAs system for
x > 0.3. Therefore, the CV measurement at 77 K probably underestimates the shallow
donor density, and overestimates the DX center density.
4.3. Summary
The electrical characterization of the Sn and Tedoped AlxGa_xAs (0.3 < x < 0.4)
grown by LPE technique has been carried out using the combined DLTS, TSCAP, constant
temperature CV and Hall effect measurements. The combined TSCAP and CV measure
ments enable the accurate determination of the densities of DX center and shallow donors
in AlxGalxAs for x > 0.3. For x > 0.3, the dominant carrier freezeout occurs and accord
ingly the accurate determination of DX center density becomes impossible by the DLTS
measurement. The ratio of Hall to drift mobility due to multivalley conduction should be
incorporated in the calculation of total electron concentration at room temperature. This
ratio is multiplied to the Hall electron concentration determined by the Halleffect measure
ment, and the total electron concentration is then determined. Both the shallow donors
and the DX center comparably supply the free electrons to the multi conduction bands of
the AlxGalxAs at room temperature.
Table 4.1. Defect parameters, shallow donor and free carrier densities
determined by the DLTS, TSCAP, CV, and Halleffect measurements.
Sample ND (cm3) NSD n (at 300 K) EDX (eV) NDX
Alo.33Gao.67As 5.91E16 1.74E16 Ec0.20 3.58E16
(Sndoped)____ Ec0.30 5.95E15
Alo.4Gao.6As 9.95E16 9.2E15 Ec0.20 7.44E16
(Sndoped) Ec0.30 1.59E16
Alo.38Gao.62As 1.54E17 1.3E16 2.2E16 Ec0.30 1.41E17
(Tedoped)
CHAPTER FIVE
PROTON AND ELECTRON IRRADIATION INDUCED
DEEP LEVEL DEFECTS IN ALGAAS
5.1. Introduction
Particle irradiation such as ion, electron, proton, and neutron induces the defects in the
active region of GaAs/A1GaAs device. Irradiation can takes place, intentionally or uninten
tionally, during the growth process or in the operating condition. Ion implantation is now
an established method for doping control in the GaAs device, for which a diffusion doping
technology has not been developed. Ion implantation doping offers a number of advantages
in the fabrication of GaAs/AlGaAs devices such as independent control of the doping level,
its depth profile, uniformity and reproductibility of doping. However, radiationinduced
defects are observed to produce effective compensating centers which are stable at room
temperature and convert the doped GaAs layer into a semiinsulating layer [4347]. An
nealing steps are needed to remove the compensating center and to activate the doping
effects.
The concentration and distribution of irradiationinduced defect can be controlled to
some extent. The defect concentration is proportional to the dose of irradiation; the distri
bution of defect is the function of the irradiant particles, their energy, and the impurities
contained in semiconductor; the impurities have the abilities to trap the intrinsic defects
originally produced by irradiation.
We can compare the energy spectrum of defects introduced by neutron, proton, and
heavy ion irradiation. In general, one expects heavier particles to create more complex
damage; either closely spaced point defects which mutually perturb each other's energy
levels, or extended defects where point defects totally lose their identity [48]. In terms
of the DLTS spectra, it is the general trend that as the mass of the highenergy particle
increases, the broader and deeper DLTS spectrum is shown. The lowfluence proton damage
is quite similar to 1 MeV electron damage, with a general trend towards a relatively larger
proportion of the damage being the E4 (Ec 0.76 eV) and E5 (Ec 0.86 eV) levels.
The Tedoped A10.38Gao.62As cells grown by LiquidPhaseEpitaxy were irradiated by
300 keV, 1 MeV proton with fluences of 1011 cm2, 1012 cm2. The 1 MeV electrons were
also irradiated onto the Tedoped Alo.38Gao.62As with fluences of 1015 cm2, 1016 cm2.
The irradiationinduced defects were investigated by the DeepLevelTransientSpectroscopy
(DLTS) and ThermallyStimulatedCapacitance (TSCAP) methods.The measured defect
parameters were correlated with the results of CurrentVoltage (IV) and Capacitance
Voltage (CV) methods.
5.2. Theoretical Review of Defect Creation by Particle Irradiation
5.2.1. Dynamics of Collision
The incident energetic particle irradiated into the solid lattice interacts with the elec
trons and the nucleis. It loses its energy by several processes depending on the nature of the
particle and its energy. The defects are usually introduced by the collisions of particles with
the nucleis which cause the atomic displacements. The concentration of produced defects
per unit time, v, is expressed by
v = aN (5.1)
where a is the cross section which is characteristic of the interaction leading to atomic
displacements; N is the nuclei concentration per unit volume; 0 is a flux of particles. The
flux, 0 is defined by
S= nv (5.2)
where n is the particle density and v is the velocity of incident particle. The probability
that the particles penetrate and make collisions with nucleis at the distance, x, is given by
the exponential term
p(x) = exp(oNx) (5.3)
Then, the average penetration depth of the particle is defined as
S= fo xp(x)dx (5
fo p(x)dx
The figure 5.1. shows the dynamics of elastic collision. Let E be the kinetic energy of an
incident particle of mass, m. It strikes an atom (of mass M) of a solid. The kinetic energy
T transmitted to an atom depends directly on the angular deflection, 0, of the incident
particle The conservation of kinetic energy and the conservation of momentum must
fulfilled. Then, the ratio of transmitted energy to particle energy is expressed by [49,50]
T M 1 1(0)
S= 2 (5.5)
E m (1+ M/m)2
where cos0 is given by
1 + (M/m)7?
cos0 = + (M/m)+ / (5.6)
/1 + 2(M/m) + (M/m)2
Then the maximum energy Tm is transmitted for 0 = 0 such that
4Mm
T 4Mm E. (5.7)
(M + m)2
For a neutron, whose mass m is very small compared to M, the transmitted energy is given
by [49]
2m
T = E(1 cosO) for 0 < 0 < r. (5.8)
For an energetic electron (m < M) relativistic corrections are necessary [51] and the kinetic
energy transmitted to the nucleus can be expressed by
T = Tmsin2 (5.9)
where the maximum transmitted energy is given by
m E
Tm = 2 IE(2 + E (5.10)
i mc2
V(m)
m E 0
   
'4
V(M)
Figure 5.1. Dynamics of the collision between an incident particle of mass m
and energy E and an atom of mass M.
5.2.2. Differential Scattering Cross Section
In the case of irradiation of a solid by energetic particles, the interaction between the
nuclei of the solid and the incident particles takes place. The differential scattering cross
section is a function of the interacting potential. The elastic scattering is not adequately
described by the total scattering cross section. The differential scattering cross section da(O)
is more adequate to describe the elastic scattering because the angular distribution of the
scattered particles is necessary. For charged particles the interaction can be represented by a
Coulombic potential (Rutherford scattering); for neutral particles (neutrons) the interaction
is rather equivalent to collisions between rigid spheres (hardcore potential). For the hard
core potential, the differential scattering cross section is given by
da = 27(Ri + R2)2cosOsin0 dO (5.11)
where R1 and R2 represent the radii of two rigid shares. For the Rutherford potential, the
differential scattering cross section is given by
irZ2e4 MdT
da mT (5.12)
Em T2
where Z represents the atomic number of lattice atom.
5.2.3. Primary Displacements
The number of total displacements produced by the incident particles is given by the
integration of differential cross section
Tm da
a(E) = dT dT. (5.13)
This integration must be carried out independently for each type of irradiating particles
since da/dT depends on the type of interaction and the nature of the particle.
For heavy charged particles (ions), the number of displacements is given by
SrZZe4 M1 1 1
a(E) = ) (5.14)
E M2 Td Tm
where Td is defined as the threshold energy. The minimum energy to displace an atom from
its substitutional site is called the threshold energy.
For neutrons the interaction is described by the hard sphere collision and the number
of displacement is given by
o(E) = 7r(R2 + R )2(1 Td). (5.15)
1 2 Tm
For electrons, electrons must have a considerable energy (in the range of 0.1 10 MeV)
to transmit energy of order of threshold energy, Td because they have very small mass
compared to the atomic mass (M). In this energy range, electrons are relativistic and the
number of displacements is given by [52,53]
(E) = 7( Ze m 1) (5.16)
mc2(72 1) Td
where 7 is given by
7 = 1 v2/c2. (5.17)
This equation can be approximated by McKinleyFeshbach's formula [54].
5.2.4. Secondary Displacements
A primary knockon atom, to which the energy T has transmitted by the incident
energetic particle, can in turn displace other atoms when T is large enough. The number of
secondary defects can be estimated by the cascade model proposed by Kinchin and Pease
[55]. This model assumes hardsphere collisions between atoms and the kinetic energy of a
knockon atom is shared equally with the stationary lattice atom since they have the same
mass. The probability of displacements in cascade is expressed by
T
p(T) = T for T > 2Td. (5.18)
For the range of Td < T < 2Td, It is obvious that only one displacement can take place.
Thus the number of secondary displacements is given by the cross section
Tm T do
o(T) = dT. (5.19)
2Td 2Td dT
For more realistic calculations the differential cross section of displacement da can be used,
depending on the interaction potential.
If the particles have the kinetic energy of the order of Td, only single displacement
takes place and it causes the formation of vacancyinterstial pair. When Tm > 2Td, the
displacement of two neighboring atoms is possible and it leads to the formation of diva
cancy. For example, the defects produced by the electron irradiation are divacancies and
their interstitials. These defects are, however, difficult to observe because vacancies and
interstitials are mobile at low temperature; they can be observed as the complex of them
selves or complex with impurities. These defects are uniformly distributed in depths of the
order of ym with highenergy electron irradiation.
For particles which transmit very high energy to the lattice atoms compared to Td, the
total cross section for displacement can be expressed by
a = up + as (5.20)
where ap and as denote the cross sections for the primary displacement and for the secondary
displacement, respectively.
5.2.5. Formation of Amorphous Layer by Irradiation
The number of displacements produced by a single primary collision can be high if the
incident particle has the kinetic energy higher than the threshold energy. Then a cascade of
displacements .occurs and the region containing a large concentration of vacancies, i.e., the
heart of cascade, is formed and surrounded by a region containing interstials. The heart
of cascade contains divacancies ,trivacancies, etc. The irradiation is usually performed at
temperature above which interstitials and vacancies are mobile; therefore, after irradiation
the interstitials disappear and the size of the heart of cascade decreases due to the vacancy
outdiffusion. The heart of cascade is now surrounded by the point defects formed by the
association of the escaping vacancies with impurities such as E or A center. The inner part
of the cascade heart is composed of vacancy clusters since moving vacancies are agglomerate.
When a material is uniformly irradiated by ions, isolated damaged regions are first
created. Depending on the size of the damaged regions and of the dose of irradiation, these
regions start to overlap until a continuous disordered layer is formed. The disorder is found
to increase as a function of dose until it saturates at the critical dose [56]. This critical dose
depends on several parameters, in particular, (1) the nature and energy of the incident ions
and (2) temperature, because the size of the disordered region is the direct function of the
nature and the energy of the incident ions and vacancy annealing occurs. When the critical
dose is reached, the disordered layer is often said to be amorphous.
5.2.6. Range of Particle
Particles move through the solid and atomic displacements occur along their path.
They are slowed down due to the energy transfer to the electrons of the solid. This result
causes the ionization of atom. For the electronparticle scattering the rate of energy loss,
or stopping power, dE/dx can be written in terms of the differential cross section do(E,T)
[57]
dE Tm
dx Ne T da(E, T) (5.21)
T [ITMrnin
where Ne is the density of electrons in the solid and Tnn and Tm are the minimum and
maximum energy transferable to an electron (of mass m) by the incident particle (of mass
M and energy E), respectively. The range R(E) of the particle is obtained directly from
this stopping power;
oE dE
R(E) = (5.22)
(dE/dx)"
The minimum energy Tm is usually considered to be the ionization energy of an electron from
its atom because energy transfer smaller than the ionization energy does not appreciably
contribute to the slowdown of the incident particles. For incident charged particles, the
Rutherford scattering law gives
dE f 4((/M)E TrZ2e4 M dT
d = E mT2' (5.23)
that is
dE rZ2e4 M 4mE
Tx = NIE mn( (5.24)
dx Em MI
where I is the ionization energy of atom. Actually, the ionization energy of atom is different
depending on the electron states.
The notion of range of a particle is particularly important for ion irradiation since
it determines the depth distribution of the implanted ions. Because of the fluctuation in
energy loss and scattering angle at each collision, a range must be defined in terms of
probability distribution of penetration depth. The important parameters are;
1. the projected range Rp, i.e., the projection of the total range on the incident direction
of the ion,
2. the range straggling, ARp, i.e., the mean square fluctuation in range which charac
terizes the depth distribution of implanted ions in the direction of implantation,
3. the range straggling perpendicular to the direction of implantation.
The numerical calculation has been performed by Lindhard, Scharffand Schrott [58]. Details
of the theory of ion range is available [59].
5.3. Results and Discussion
The proton and electron irradiations produced different defect states and the different
energy of irradiant rendered the variation of final defect state. The observed defect states
were the Ec 0.28 eV, Ec 0.39 eV, Ec 0.55 eV, Ec 0.76 eV, Ec 0.86 eV and Ec 0.99
eV levels as are shown in Fig. 5.2., 5.3., 5.4., 5.5., 5.6., and Fig. 5.7. No observable hole
trap levels were detected. The Ec 0.28 eV level is known as DX center and this level exists
dominantly both in the irradiated and unirradiated A10.38Gao.62As cells in terms of defect
density. This level (E3) was observed in the irradiated GaAs but the measured activation
energy is widespread from 0.31 eV to 0.45 eV [6062].
The Ec 0.55 eV level is very stable at room temperature. This level was observed in
the Alo.38Gao.62As irradiated by 300 keV proton with the fluence of 1012 p/cm2 and in the
one irradiated by 1 MeV with the fluence of 1016. This level was observed only in the device
of high dose. It is definitely irradiationinduced level. This level was reportedly observed in
the Be+implanted ntype AlGaAs and it was assigned as the Becomplex [63]. It is possible
that the irradiation induced vacancies are associated with Be impurities which are used as
the ptype dopant in this sample.
It should be noticed that the Ec0.76 eV and Ec0.86 eV, which seem like EL2 group,
were observed. The Ec0.86 eV level was also observed in the preirradiated Alo.38Gao.62As
cells. The density of the Ec0.86 eV level was significantly increased after proton and/or
electron irradiation.
The difference between 300 keV and 1 MeV proton irradiation is that the 300 keV
proton irradiation produces the higher overall density of irradiationinduced defects but the
1 MeV protons induce the midgap level of Ec0.99 eV with the big capture cross section
of 1.37E9 cm2. This level probably causes the major degradation of the A10.38Gao.62As
cells after the 1 MeV proton irradiation. This level showed the long tail of emission from
low temperature and the broad DLTS spectrum as shown in Fig. 5.2. and in Fig. 5.6.
The similar phenomena was observed in the neutronirradiated AlGaAs [64]. The neutron
induced level (assigned as the defect clusters) showed the almost same activation energy as
the Ec 0.99 eV and the long tail of broad DLTS spectra which comes from the enhanced
thermal emission at low temperature due to the electric field associated with the defect
clusters. The neutron and the proton have the identical mass. However, the transmitted
energy is quite different because the energy transfer to atom by protons is due to Rutherford
scattering, whereas the energy transfer by neutron is due to elastic scattering. Therefore,
the neutron irradiation can cause the heavier damage to the lattice. It is plausible that the
highenergy (1 MeV) proton irradiation gives the similar effects as the neutron irradiation
1 MeV Proton
Fluence: 1011 cii2
Ec 0.99 eV
e, = 34.4 S1
Ec 0.28 eV
200
250
300
TEMPERATURE (K)
Figure 5.2. The DILTS spectra of the Tedoped A10.3asGC .62As irradiated
by 1 MlV proton with fluence of 10o" c(112
100
150
350
_ __ ___ __
_ _~__ __ ___ __
1 MeV Electron
Fluence: 1010 cm2
Ec 0.55 eV
en = 34.4 s1
o
1001
t>
!ii
*+1
Ec 0.28 eV
150 200 250 300 350
TEMPERATURE (K)
Figure 5.3.
The DLTS spectra of the Tedoped Alo.38Gao.62As irradiated
by 1 MeV electron with fluence of 1016 cm2.
EC 0.86 eV
V~_L\~I**/pV\~M/.v
Ec 0.86 eV
en = 34.4 s1
1 MeV Electron
Fluence: 1015 cm2
4o
(Jd
>
rtl
L
100
150 200 250 300 350
TEMPERATURE (K)
The DLTS spectra of the Tedoped Alo.38Gao.62As irradiated
by 1 MeV electron with fluence of 1015 cm2.
EC 0.28 eV
Figure 5.4.
300 KeV Proton
Fluence: 1012 cm2
Ec 0.55 eV
EC 0.86 1
SEc 0.76 eV
Qen = 34.4 s1
I I I I
200 250 300 350 401
TEMPERATURE (K)
Figure 5.5. The DLTS spectra of the Tedoped Alo.38Ga0.62As irradiated
by 300 KeV proton with fluence of 1012 cm2.
0
eV
Cz
L
200
Figure 5.6.
The DLTS spectra of the Tedoped Alo.38Gao.62As irradiated
by 1 MeV proton with fluence of 1012 cm2
250 300 350 400
TEMPERATURE (K)
300 KeV Proton
Fluence: 1011 cm2
U) e,, = 34.4 s1
SEc 0
Ec 0.28 eV
100 150 200 250 300
TEMPERATURE (K)
Figure 5.7. The DLTS spectra of the Tedoped Alo.sGao.62As irradiated
by 300 KeV proton with fluence of 1011 cm2.
).76 eV
L50
if the fluence and the energy of protons is increased. It must be noted that the Ec 0.99
eV level is not observed in the lowenergy proton irradiated A1GaAs and in the electron
irradiated one. However, its density is quite small and it is in the order of 1014 cm3 in the
highenergy proton irradiated samples. The estimated lifetime (7n = aooNTVth) is in the
order of 1013 sec but the real lifetime rn = anNtvth is in the order of 1011 second if the
multiphonon emission is considered as an = a, exp(E).
The CV measurement at room temperature will give the total ionized impurity density
which composes of the sum of shallow level and DX center densities since the DX center at
the edge of spacecharge region follows the 1 Mhz small signal and contributes to the mea
sured junction capacitance. The total free carrier at the conduction bands were calculated
by [65]:
n = ND NT (5.25)
The IV measurement showed that ideality factor is in the range of 2.16 to 2.55. The device
# 43 (300 keV proton, fluence of 1011 cm2) has the ideality factor of 2.16 to 2.25. The
device # 44 (300 keV proton, fluence oflO12) has the ideality factor of 2.55. The device
# 46 (1 MeV proton, fluence of 1011 cm2) showed the large variation of 2.20, 2.26, and
2.53 in ideality factor from sample to sample. The device #47 (1 MeV proton, fluence of
1012 cm2) has the ideality factor of 2.22. The deveces # 41, 42 (1 MeV electron, fluences
of 1015, 1016 cm2) have the ideality factors of 2.21 and 2.22, respectively. The measured
defect parameters were summarized in the Table 5.1.
Table 5.1. High Energy Proton and Electron Irradiation Induced Defect Parameters
Determined by DLTS, TSCAP, CV, and IV Method in Tedoped Alo0.38Gao.62As Cells.
Energy Fluence nt ND ET Nt ooo rn
(MeV) (cm2) (cm3) (cm3) (eV) (cm3) (cm2) (sec)
Proton
0.3 Ell 1.3E17 1.48E17 Ec0.28 1.35E17 3.6E14 2.02E11
Ec0.76 1.10E15 8.43E14 1.08E9
E12 7E15 1.41E17 Ec0.28 1.34E17 3.66E14 2.04E11
Ec0.55 4.46E15 4.54E14 5.17E10
Ec0.76 7.99E15 8.43E14 1.48E10
Ec0.86 7.05E15 6.73E14 2.11E10
1 Ell 2.8E16 1.54E17 Ec0.28 1.26E17 3.66E14 2.17E11
Ec0.99 7.87E14 1.37E9 9.27E14
E12 3.6E16 1.41E17 Ec0.28 1.05E17 3.66E14 2.60E11
Ec0.39 7.36E14 1.24E16 1.10E6
Ec0.86 6.13E14 6.73E14 2.42E9
Ec0.99 4.91E14 1.37E9 1.49E13
Electron
1 E15 1.9E16 1.33E17 Ec0.28 1.14E17 3.66E14 2.40E11
Ec0.86 7.60E14 6.72E14 1.96E9
E16 3.6E16 1.63E17 Ec0.28 1.27E17 3.66E14 2.15E11
Ec0.55 8.54E14 4.34E14 2.70E9
Ec0.86 1.16E15 6.73E14 1.28E9
0 2.5E16 1.05E17 Ec0.28 8.0E16 3.66E14 3.42E11
Ec0.86 5.67E14 6.73E14 2.62E9
CHAPTER SIX
GROWNIN DEFECTS IN TEDOPED ALxGAixAS GROWN
BY LIQUIDPHASEEPITAXY VS GROWTH TEMPERATURE
6.1. Introduction
The compound semiconductor, AlxGalxAs, is widely used for the optoelectronic de
vices such as double heterostructure lasers and heterface solar cells prepared by liquid
phaseepitaxy technique. Recently, much interest has been focused on this material because
of its application in fabricating highly efficient photovoltaic devices. In particular, multi
junction cascade solar cells using the GaAsAlGaAs system have been reported to have a
potential efficiency of as high as 30 % at high solar concentration [66]. For the fabrication on
ptype A1GaAs, Zn and Ge have been commonly used. However, they have the shortcom
ings as follows. Zn is not suitable for the ptype dopant because of its high diffusivity for
those multijunction structures, in which sharp impurity profiles are required and additional
epilayers have to be grown on the top of the Zndoped layer [67]. Furthermore, the high
vapor pressure of Zn causes serious contamination in growing multilayers in an open tube
liquid phase epitaxial (LPE) system. Ge has shown to be difficult to obtain high carrier
concentration in AlxGaxlAs with x > 0.3 because of its strong dependence of the acceptor
ionization energy on the Al composition and its compensation effect due to amphoteric
behaviour [68,69]. In contrast, Be has the desirable characteristics such as low diffusion
coefficient, low vapor pressure, and high distribution coefficient.
The behavior of Ge and Zn as a dopant in AlxGalxAs has been well investigated. It
is wellknown that besides the presence of the dopant related centers (DX centers), deep
hole traps have been detected at both sides of pn junction. The physical origin of such
hole traps, present in moderately large concentrations, is discussed in terms of Znrelated
or Gerelated complexes located at the Ev + 0.40 eV [7074]. However, very few Berelated
acceptor levels have been reported in the literature.
6.2. Experimental
Tedoped AlGaAs homojunction mesa diodes grown by LPE have been used in this
study. The Al composition has been controlled by the growth temperatures and the char
acteristics curve of Al composition versus growth temperature are shown in Fig. 6.1. The
relationship between the Al composition and growth temperature is linear. This relationship
was used for the determination of Al composition in LPE AlxGalxAs. Growth tempera
ture is in the range of 7020 C to 7800 C and it corresponds to the Al composition range of
0.23 to 0.41. The substrate used for the mesa A1GaAs homojunction is n+ GaAs doped by
Si. The samples of #6 and #7 have the same growth temperature but the device #7 has an
undoped AlxGalxAs buffer layer between n+ GaAs substrate and Tedoped AlxGalxAs
active layer as shown in Fig. 6.2. The epilayer thickness varies from 2.4 to 5.4 um and
the device area is 4.153 x 104 cm2. The capacitancevoltage technique was used for the
determination of background doping concentration. The IV measurements were used for
the determination current component under low forwardbias condition. The DLTS mea
surements were performed in order to determine the physical parameters of grownin defects
in AlxGalxAs homojunction devices.
6.3. Results
One dominant electron trap (DX center) and two hole traps were observed from the
DLTS measurements. The results are summarized in Table 6.1. The Terelated center (DX
center) has an activation energy from 0.28 eV to 0.31 eV below the conduction band. The
density of DX center increases as Al composition increases but near the crossover point of
directindirect bandgap it becomes saturated and slightly decreases. This results generally
agrees well with the reported data indicating that the concentration of DX center decreases
as Al composition goes far beyond the directindirect bandgap crossover point [75].
1.0
0.8
0
O
V)
0
O
0.6
0
S0.4
0.2 
700 720 740 760 780 800
TEMPERATURE (oC)
Fig. 6.1. Characteristic curves of Aluminum composition versus growth temperature.
Device # 1, 2, 3, 4, 6
p+(Be)AlxGalxAs
n (Te)AIlGai_,As (2.4 5.4 pm)
n+(Si)GaAs
Device # 7
p+(Be)AlxGaixAs
n (Te)AlxGaixAs (2.4 pm)
undoped AlxGaixAs
n+(Si)GaAs
Figure 6.2. Test device structure of LPE grown AlxGalxAs mesa diodes.
As for the hole traps, their activation energies become deeper as the Al composition
increases. The shallow one (Ev + 0.25 to 0.28 eV) has the almost comparable densities
(about 4 x 1016 cm3) except for device #7, which has the undoped AlGaAs buffer layer
between substrate and active layer. The device #7 has no considerable hole trap at around
0.25 eV above the valence band. It has been reported that in Sndoped GaAs grown by
liquidphaseepitaxy under arsenicdeficient conditions, compensation due to the possible
occupation of As sites by Si may exist [76,77]. A broad photoluminescence band is observed,
and it suggests that the Si acceptor level is located at the Ev + 0.20 eV. Therefore, the
measured value of the Ev + 0.25 to 0.28 eV is likely to be reasonable, assuming that the
Sirelated acceptor level becomes deeper as the Al composition increases. It looks likely
that the buffer layer deters the diffusion of Si from the n+ GaAs substrate which is heavily
doped by Si. In comparison of #6 and #7, it is clear that the buffer layer slows down the
diffusion of Si from the substrate. The new level at the Ec 0.32 eV looks like the Sirelated
DX center. Otherwise, another Terelated DX center is a possible candidate because Te and
Se can have two DX center in the Al composition range of 0.3 to 0.6. Actually, every sample
showed a small step at the right hand side of DLTS spectrum of the Terelated DX center
suggesting the presence of another level of smaller density.
As for the deeper hole trap, it becomes deeper with increasing the Al composition;
its activation energy varies from Ev + 0.68 eV to Ev + 0.93 eV with corresponding Al
composition range of 0.27 to 0.41 as shown in Fig. 6.3. In Alo.23Gao.77As, however, this
level was not observed. It is worthwhile to review the behaviour of other ptype dopant
materials such as Zn and Ge in AlxGalxAs for the purpose of comparison. Ruling out
occasional contamination, there is strong evidence that for Al compositions near bandgap
crossover, donors and acceptors tend to form both shallow and deep centers [78]. In fact,
dopants incorporate significantly into the deeper positions as aluminium content is close to
0.4 and goes beyond 0.4. Ge introduces two shallow acceptors and one deep acceptor in LPE
.0
Ec 0.31 eV
I I I I
150 200 250 300
TEMPERATURE (K)
The DLTS spectra of Alo.31Gao.69As grown by Liquid Phase Epitaxy.
Figure 6.3.
AlGaAs, with the dominant one of larger activation energy increasing its activation energy
to 160 m eV at x = 0.4 [70,71,79]. It has been observed that Ge introduces Gecomlex at
about 0.4 eV from the valence band [70,71]. Also, for Zn in AlxGalxAs grown by LPE and
by MOCVD, the presence of a Zncomplex center (proposed as ZnGaVAs) at the Ev + 0.4
eV was confirmed, only with the Al composition range of near or beyond the directindirect
bandgap crossover point with its density strongly depending on the Al composition [7274].
This fact qualitatively agrees with our results. The deeper hole trap at the Ev + 0.68
eV to 0.93 eV ascribed to Berelated complex (proposed as BeGaVAs) were observed only
in the Al composition range of 0.27 to 0.41, not in A10.23Ga0.77As. Also, its density has
a large variation from sample to sample; the device #3 of Alo.31Gao.69As has a relatively
small density of 1.42 x 1015 cm3 whileas the device #4 of Alo.27Gao.73As has the highest
density of 2.12 x 1017 cm3. One of the possible explanations is that at about x = 0.31
the Berelated complex (BeGaVAs) yields the smaller density due to the unknown reasons
compared as the samples of x = 0.41, 0.36, and 0.27. An almost comparable observation
was reported by Calleja et. al. in Zndoped LPE AlxGalxAs [77]: The sample of Al
composition of 32 % showed the smaller concentration of the Ev + 0.40 eV center with the
highest external quantum efficiencies as compared to other samples. This fact qualitatively
agrees with our IV measurements in which the sample of x = 0.31 has the best ideality
factor as compared to other samples of 0.27 < x < 0.41. Generally, the results of IV
measurements are consistent with the data of DLTS measurements. For the samples #6
and #7, there exists a considerable discrepancy in the ideality factor probably due to the
twodimensional conductivity in device #7 at the interface between undoped AlxGalxAs
and substrate.
6.4. Conclusions
In Tedoped LPE AlxGalxAs homojunction diode, the grownin defects have been
investigated by the DLTS, IV, and CV measurements. The grownin defects are related to
the n or ptyped dopant materials. The density of DX center increases as the Al composition
increases. However, near the crossover point of directindirect bandgap it becomes slightly
decreased. Two hole traps have been observed related to the Be and Si dopant materials
which diffuse from the p+(Be) AlxGalxAs layer and from the n+ (Si) GaAs substrate,
respectively. The BeGaVAs complex has shown the characteristics similar to the ZnGaVAs
and GeGaVAs complexes in LPE and MOCVD AlxGalxAs. The BecaVAs complex has
been detected only in the Al composition range of 0.27 < x < 0.41, not in Alo.23Gao.77As.
This result is consistent with the features of Zn and Gerelated complexes in AlxGaixAs.
At x = 0.31, the IV characteristics has shown the best ideality factor of 2.06 among those
samples in the range of 0.27 < x < 0.41.
Table 6.1. Defect parameters determined by DLTS, IV and CV measurements.
Device ND(cm3 ET(cm3) NT(cm3) an,p(cm2) Possible origins
#1 2.40E17 Ec0.30 1.65E17 3.41E14 DX center(Te)
T=7600C Ev+0.26 4.08E16 5.56E16 Sirelated
x=0.36 n=2.07 Ev+0.88 1.88E16 7.13E13 BeGaVAs
#2 "2.18E17 Ec0.28 1.00E17 1.09E14 DX center(Te)
T=7800C Ev+0.28 3.60E16 5.86E15 Sirelated
x=0.41 n=2.11 Ev+0.93 4.27E16 1.16E11 BeGaVAs
#3 2.13E17 Ec0.31 1.63E17 2.52E13 DX center(Te)
T=7400C Ev+0.25 4.39E16 1.49E15 Sirelated
x=0.31 n=2.06 Ev+0.73 1.42E15 1.32E14 BeGaVAs
#4 2.95E17 Ec0.31 3.12E16 6.02E13 DX center(Te)
T=7210C Ev+0.25 5.55E15 5.56E16 Sirelated
x=0.27 n=2.13 Ev+0.68 2.12E17 8.32E15 BecaVAs
#6 3.41E17 Ec0.28 2.23E16 3.12E14 DX center(Te)
T=702C Ev+0.24 4.72E16 2.60E13 Sirelated
x=0.23 n=1.64
#7 3.97E17 Ec0.31 1.58E16 9.16E13 DX center(Te)
T=702C Ec0.33 1.26E16 1.01E13 DX center(Si)
x=0.23
CHAPTER SEVEN
DEFECT CHARACTERIZATION AND ELECTRICAL PROPERTIES
OF SIMOX BASED SOI DEVICES
7.1. Introduction
Recent development in SiliconOnInsulator (SOI) technology has shown promise for
providing a viable technique for the fabrication of fully isolated transistor structures. Sev
eral SOI approaches have been attempted. Among them, the Separation by IMplanted
OXygen (SIMOX) is a promising technology for complementary metaloxidesemiconductor
large scale integration. In SIMOX technology, a highdose of oxygen ion implantation is
applied to the silicon substrate at high temperature. The effect of oxygen implantation
on the top silicon layer over the buried oxide is the main topic of research interests and
concern for the SIMOX based SOI devices. Since the top silicon layer is heavily damaged
during oxygen implantation, annealing process is used in all SIMOX based materials and
devices. The morphologies of the silicon/buried oxide/substrate have been studied using
various experimental methods such as Rutherford backscattering spectroscopy (RBS) and
crosssectional transmission microscopy (XTEM). However, very few experimental data
concerning the electrically active defects in the SIMOX materials and devices have been
published by using the DLTS method. The DLTS technique is a powerful tool for direct
observation of electrically active defect levels in SOI materials and devices induced by the
oxgen implantation and the subsequent annealing process. The IV and CV methods can
also be applied for electrical characterization of the SIMOX devices and their respective
results can be correlated to the DLTS results.
7.2. Fabrication Process
The buried oxide structure was produced by using 150 KeV 0+ ion implantation onto
the 35 ohmcm ntype silicon substrate. The substrate temperature was kept at 500 OC.
The resultant buried oxide layer has a thickness of about 0.4 pm. The superficial silicon layer
is heavily damaged due to the 0+ ion implantation and hence annealing process should be
used in order to recrystallize the damaged top silicon layer over the buried oxide. This step
is a prerequisite for the silicon epilayer growth. Different annealing conditions were applied
to study their effects on the defects and electrical characteristics of the SIMOX based SOI
devices. Annealing temperature was varied between 1150 and 1350 o C and annealing time
varied between 2 and 24 hours. The dopant density of n epilayer was 1 x 1015 cm3 and its
thickness varied between 0.8 and 2.5 pm. The pwell region was formed by boron implant
and subsequently drivein diffusion into the n epilayer. The doping density of pwell region
is 2 x 1016 cm3. The p+/n and n+/p junctions were formed by implantation of boron and
phosporous, respectively. The p+/n diodes were used to study the defects and electrical
properties of the nepilayer and the n+/p diodes were used to study the defects in the pwell.
The lateral isolation was incorporated in batch #3 and #4 SOI devices by using LPCVD
growth of thermal oxide and undoped polysilicon. The test structure is shown in Fig. 7.1.
The process parameters of batch #1, #2, #3, and #4 are listed in Table 7.1.
7.3. Oxygen Related Defects in SIMOX Based SOI Devices.
The DLTS measurements on batch #1 through #4 SIMOX devices revealed various
defect levels. The measured defect parameters and the tentative assignment for the origins
of these defects are summarized in Table 7.2 and Table 7.3. The Ev + 0.65 eV hole level
was observed in all the devices studied. It is anticipated that the oxygen implantation
and the subsequent annealing processes will generate the oxygenvancancy complex and
other related defects in the SIMOX based SOI devices. The implantationinduced vacancies
DEVICE STRUCTURE
Lateral Isolation
Lateral Isolation
(1) Epi Thickness; 2.0 pm
(2) Oxide Thickness; 0.4 pm
(3) n epi; 3 x 1015 cm3
(4) p; 2 x 1016 cm3
(5) Substrate; 3 5 ohmcm
Figure 7.1. Test structure of SOI device fabricated by SIMOX
(Separation by Implant of Oxygen) technique.
enhance the impurity diffusion. Therefore, the creation mechanism for the impurity related
defects is rather complex in the fabrication of SOI devices.
Aside from the stable defects observed in the tested SOI devices, some anomalously
unstable defects which are probably due to the mobile ions such as alkali metal impurities
were also observed in some SIMOX devices with lateral isolation (batch # 3 and #4).
The unstable defect levels were also observed in the control devices of batch #3 and #4,
indicating that unstable defect levels may be attributed to the metallic contamination during
lateral isolation and other high temperature processes applied during the device fabrication.
The oxygen ion implantation onto the silicon substrate creates the buried oxide layer
and induces vacancy and oxygen related defect clusters in the heavily damaged silicon
surface layer. Prior to the high temperature annealing, the top silicon layer and the substrate
region adjacent to the buried oxide layer were heavily damaged. It has been shown that a
three hour annealing at 1150 OC on the SIMOX processed silicon film, the damaged surface
layer will be recrystallized to form a single crystal silicon [80]. High densities of oxygen
precipitates and dislocations can be observed in both silicon film and the substrate before
annealing [81,82]. When the annealing temperature increases to 1250 OC or higher, the
entire top silicon layer becomes precipitatefree [80,83]. The oxygen content in the single
crystal region is at least an order of magnitude higher than the solid solubility limit [84].
During high temperature annealing, possible sinks for the supersaturated oxygen atoms are
the oxygen precipitates, the surface and the top silicon/buried oxide interface.
In the SIMOX based SOI devices studied an anomalous capacitance step was observed
in the CV curve of the p+/n junction diode formed on the nepilayer. This capacitance (a
reduction of net dopant density) step may be due to the electricfield shielding effect of the
oxygen precipitates [85,86]. This phenomenon seems unique in the SIMOX devices studied,
which occurs when the depletion edge reaches the epi/oxygenprecipitates or the epi/oxide
interfaces. This phenomenon was not observed in the pwell region or in the control devices.
The reason that this anomalous capacitance step was not observed in the pwell region can
be explained as follows. The pwell fabrication step consists of boron implantation and
subsequent drivein process at 12000C. Since boron diffusion is much faster than oxygen,
the pwell region is not much encroached by the diffusion of oxygen. Another explanation
for this capacitance step shown in the CV plot of SIMOX devices is due to the increase
of positive fixed charge at the epioxide interface. The observed anomalously unstable
defect levels by the DLTS measurements are attributed to the contamination of alkali metal
impurities during the lateral isolation process. However, the increase of positive oxide fixed
charge actually reduces the width of depletion region next to the oxide layer in ntype
material. Thus the increase of positive fixed oxide charge will make the capacitance step
in the CV plot occurred closer to the epilayer/buried oxide interface in the ntype epilayer.
This fact is contrary to the observed results of batch #3 and #4 SIMOX devices. The
presumption that anomalous capacitance step is due to the extension of oxide depletion
region into the nepilayer is questionable. It is noted that the capacitance step occurs closer
to the surface if higher annealing temperature and longer annealing time are used, as is
shown in Fig. 7.9. This can be explained by the oxygen outdiffusion from the buried
oxide layer. For examples, for devices from S16 and S19, the annealing time (2 hrs) is
shorter than from S12, and their capacitance step was found closer to the surface due to
the thinner epilayer. The main difference observed in the SIMOX devices with and without
lateral isolation is that the capacitance step is closer to the surface for devices with lateral
isolation, as is shown in Fig. 7.10. Even if annealing temperature and annealing time were
identical, the devices with lateral isolation showed that the capacitance step was closer to
the surface as compared to the devices without lateral isolation. In fact, for devices from
batch # 3 and # 4, which have lateral isolation, additional defects and the oxygen related
precipitates may come from the thermal oxide and LPCVD polyslicon used in the lateral
isolation. A comparison of the SIMOX devices with control devices shows clearly that the
Batch # 1
S 10 p+n
ET = Ec 0.29 eV an = 1.40 x 1014 cm2
NT/ND = 6.30 x 102
150
200
250
300
TEMPERATURE (K)
Figure 7.2. The DLTS spectra of the nepilayer of SIMOXbased SOI p+/n
junction diode (batch #1).
100
350
I I I
Batch # 1 S 10 p+n
ET = Ec 0.33 eV
a,, = 2.00 x 1015 cm2
NT/ND = 1.73 x 10'
200 250 300
TEMPERATURE (K)
The DLTS spectra of the nepilayer of SIMOXbased SOI p+/n
junction diode (batch #1).
350
100
150
Figure 7.3.
p+1n
Batch # 2
S1&S4
n+p
EA = Ev + 0.55 eV, op = 3.94 x 106 cm2
EB = Ev + 0.50 eV, op = 1.29 x 1017 cm2
Ec = Ev + 0.63 eV, up = 2.91 x 1016 cm2
I I I L
100
150
200
250
300
350
TEMPERATURE (K)
Figure 7.4. The DLTS spectra of the nepilayer and pwell of SIMOXbased SOI
devices (batch #2).
100 150 200 250 300 350
Figure 7.5.
TEMPERATURE (K)
The DLTS spectra of the nepilayer of SIMOXbased SOI device
(batch # 3).
S 2 n+p
Batch # 3
t\
en = 34.4 s1
Ev + 0.65 eV
Up = 1.35 x 1016 cn12
NT/NA = 2.91 x .102
I I I I I
100 150 200 250 300 350
TEMPERATURE (K)
Figure 7.6. The DLTS spectra of the pwell region of SIMOXbased SOI device
(batch # 3).
p+n
Batch # 4
Ec 0.20 eV
al, = 1.28 x 1018 cm2
NT/ND = 0.10
200
250
300
TEMPERATURE (K)
The DLTS spectra of the nepilayer of SIMOXbased SOI device
(batch # 4).
100
150
Figure 7.7.
350
I I I I 
Ev + 0.13 eV
op = 2.1 x 1018 cm2
NT/ND = 7.2 x 102
Batch # 4
S 16
p+n
eV NT/ND = 1.2 x 102
x 1018 cm2
100 150 200 250 300 350
Figure 7.8.
TEMPERATURE (K)
The DLTS spectra of the nepilayer of SIMOXbased SOI device
(batch # 4).
VR (Voltage)
The CV characteristics of the nepilayer of SIMOXbased SOI
device (batch # 3).
Figure 7.9.
0 5 10
VR (Voltage)
Figure 7.10. The CV characteristics of the nepilayer of SIMOXbased SOI
device (batch # 4).
abrupt capacitance step observed in the CV curves of the lateral isolated SOI devices is
a direct result of the oxygenprecipitates enhanced diffusion from the recrystallyzed silicon
layer into the nepi region. The growth of lateral isolation gives additional thermal annealing
or diffusion step for the oxygen precipitates to diffuse closer to the surface region of the
epilayer. Fig. 7.10 verifies this fact.
The electrical properties (e.g., leakage current) for the SIMOX devices with lateral
isolation and the electricfieldshielding effect are believed to be closely related to the oxygen
contained in the undoped polycrystalline layer above the buried oxide [85,86].
The results of IV measurements support our assignment of the anomalous capacitance
step of the CV curve to the oxygen precipitates and oxygen related defects. Furthermore, it
was observed that the diode ideality factor for the SIMOX devices which had the capacitance
step of CV plots closer to the surface, was significantly higher. This arises from the
recombination current produced at the oxygen precipitates. The results of IV and C
V measurements are summarized in table 7.4, table 7.5, and table 7.6. The oxygen and
Si02 precipitates diffuse out to the surface or migrate close to the interface region of top
silicon and buried oxide during both annealing and lateralisolation growth process at high
temperature. In other words, the transition part of the CV curve corresponds to the oxygen
precipitates or Si02 precipitates. For example, the observed defects levels, the Ec 0.20 eV
and Ec 0.28 eV by the DLTS, are ascribed to the vacancy plus oxygen complexes, V30 and
V20, respectively, for the batch #4 devices (S6, S16) [87]. The applied reversebias pulse
was 0 to 3 V, of which the voltage range covered the anomalous capacitance step of the CV
curve. It is expected that the oxygen and vacancy complexes can be found in the region
corresponding to the capacitance step part of the CV curve in the SIMOX devices. The
location of the oxygen or SiO2 precipitates may be decided based on the prolonged annealing
process including the growth of lateral isolation and the epilayer growth at 11000C over the
recrystallized top silicon layer. Due to the difference in solubility at 1250 OC and at 875
OC [88], the silicon layer is still supersaturated with oxygen at 875 OC during the growth
of thermal oxide for the lateral isolation, and hence the oxygen precipitates may still be
formed but with a much lower density.
The final stopping location of oxygen precipitates is quantitatively predictable as a
function of effective annealing temperature, time, and epilayer thickness. Here, the effective
annealing temperature and time are defined as including the growth temperature and time of
thermal oxide and polysilicon. A computer model can in principle be developed to simulate
the SIMOX device fabrication using different process parameters.
In order to deter or slow down the diffusion of oxygen into the epilayer of SIMOX
devices, a two step annealing process at annealing temperatures of 1250 and 1150 C is
suggested. Annealing time of at least 3 hours is required at each temperature. After the
twostep annealing, the surface of silicon layer, which contains high percentage of oxy
gen precipitates after post implantation annealing, must be removed prior to the epilayer
growth. The second annealing performed at 1150 OC is required in order to getter the su
persaturated oxygen (at least one order higher than the maximum solid solubility) in the
single crystallized silicon layer over the buried oxide since the subsequent growth of epilayer
and the thermal oxide growth were processed near 1150 C. This approach will reduce the
oxygen precipitates significantly in the active region of the SIMOX device.
7.4. Correlation of Electrical Properties with Process Parameters
7.4.1. Characterization Methods of SIMOX SOI Devices
In this study, the DLTS, IV and CV methods were used to characteristize defects and
electrical properties of SIMOX devices with different process parameters. Oxygen dose, epi
thickness, and postimplantation annealing condition were changed in order to investigate
the effect of individual process parameters on the defects and electrical properties of the
SIMOX based SOI devices fabricated at Harris Semiconductor. Correlation between process
history and electrical characteristics and defects observed in these devices were discussed
for batch #1 through #4 so that process parameters can be optimized.
7.4.2. Epilayer Thickness
In batch #4 devices, the epithickness varies from 1.7 to 2.5 pm. Experimental results
show that small change in epithickness has little influence on the electrical characteristics
( except batch #3 S8 and #4 S16 ). In general, the thicker the epilayer, the better the
electrical characteristics. The variation in leakage current is too small to tell the difference
between different devices, whereas the ideality factor and breakdown voltages show good
agreement with the above statement. The only exception is for devices in slice S6 and S12.
The difference may be due to the long annealing time and will be discussed later.
7.4.3. Annealing Time
From the published papers [80], it has been shown that high temperature heat treat
ment on SIMOX materials for six hours produce a precipitatefree silicon layer. Shorter
annealing time may cause degradation of quality in Si/Si02 layer. It seems that during
epilayer growth, oxygen may diffuse into the epilayer and degrade the electrical properties.
The layer between the oxygenfree epilayer and the buried oxide may consist of a mixture
of polycrystaline silicon and silicon oxide. This layer has the electricfieidshielding (EFS)
effect and can be used to improve the breakdown voltages of MOSFETs fabricated using
SIMOX technology [81].
A comparison of batch #2, #3, and #4 reveals that the effective annealing time,
which includes postimplantation annealing time, epilayer growth time, and lateral oxide
growth time, and the process temperatures, principally decides the electrical characteristics
of the SIMOX devices. The main reason of degradation in electrical characteristics of batch
#3 and #4 devices is that the extra step for the lateral isolation causes the alkali metal
contamination and the diffusion of oxygen precipitates into the device active layer. The
significant increase of breakdown voltage of nepi of the SIMOX devices is closely related to
the formation of the electricfieldshielding layer (EFS layer) due to the oxygen precipitates.
On the other hand, the breakdown voltage of pwell diodes for the SIMOX materials is equal
to or less than the breakdown voltage of the control devices. We have already discussed
that the diffusion of the oxygen precipitates into the pwell region of the SIMOX device is
blocked due to the fast diffusion of boron from the surface of epilayer instead of oxygen
diffusion from the recrytallyzed silicon layer during the boron drivein process at 12000C.
The formation of EFS layer significantly increases the breakdown voltage of SIMOX devices.
The IV measurements are entirely consistent with the CV measurements. The SIMOX
devices of batch #3 and #4 show that the large ideality factor is attributed to the recombina
tion current component at the interface region of epi/oxygenprecipitates which corresponds
to the region where capacitance step occurred. The SIMOX devices with capacitance step
occurred closer to the surface consistently show a higher diode ideality factor.
7.4.4. Lateral Isolation
Lateral isolation is believed to be the main reason for the degradation of the IV and
DLTS characteristics observed in batch #3 and #4 devices. Incorporation of lateral iso
lation can cause great degradation in electrical characteristics, especially, leakage current
and ideality factor. Breakdown voltage is also closely related to the lateral isolation. Incor
poration of lateral isolation increase breakdown voltage significantly.
7.4.5. Ion Dose and Annealing Temperature
The similarity of electrical characteristics between batch #4 S16 and S19 shows the
compensation effect of ion dose and epithickness. The correlation between ion dose and
electrical properties still need more study. The effect of 0+ dose can be investigated more
clearly if the other process parameters such as annealing time and temperature are identical.
7.4.6. Defects and Electrical Properties of Pwell
In most SOI devices studied, the electrical characteristics of p+/n junction diodes
are superior than the n+/p junction diodes. This phenomenon can be attributed to the
contamination in the boron implanted p well. The origins of the defects appeared to be
related to foreign..impurities introduced during the implantation and drivenin diffusion
process.
7.4.7. Nepilayer/pwell Interface
The experimental results of batch #1 devices showed that in p/n junction diodes,
which were intended for characterizing the epilayer in the lateral direction, several new
traps were found. These traps may be related to the nepilayer/pwell interface related
defects.
7.5. Summary and Conclusions
1. Traps A, B, C, I, and M appeared to be introduced during the boronimplant or
drivein process since these traps were found only in pwell region.
2. Most of traps found in nepilayer have smaller activation energy than traps found in
the pwell. These traps are believed to be related to the foreign impurities introduced
during oxygen implantation or due to outdiffusion of oxygen atoms.
3. The density of most defects increases with depth in the nepilayer. It may be related
to the outdiffusion of oxygen precipitates from the buried oxide to epilayer.
4. Lateral isolation process has the most significant influence on degrading electrical
properties of SOI devices studied. This is caused by the undesirable extra annealing
time due to the lateral isolation growth. This step gives rise to the outdiffusion of
oxygen into the nepi layer.
78
5. Epithickness, ion dose, lateral isolation, annealing time and temperature are closely
related to the outdiffusion of oxygen precipitates. Optimal match between these
process parameters may be obtained by using process simulation.
6. The control of the outdiffusion of oxygen precipitates into the device active region is
the main concern for the fabrication of SIMOX based SOI devices.
Table 7.1. Process parameters of SIMOX based SOI devices.
Batch no. Slice no. O+ dose Anneal Anneal Epilayer Lateral
[cm2] Temp. [OC] Time [hrs] thickness [im] Isolation
1 S4 (simox) 1.8 to 1250 24 1.0 no
(1977BI) S5..(simox) 2.2E18 1350 2 2.0 no
S9 (simox) 1250 24 1.5 no
S10 (simox) 1150 3 1.5 no
2 S19 (control) no
(31852) S1 (simox) 1.8E18 1250 16 2.0 no
S4 (simox) 1.8E18 1250 16 2.0 no
3 S2 (control) yes
(31851) S8 (simox) 1.8E18 1250 16 2.0 yes
S15 (simox) 1.8E18 1150 3 2.0 yes
S16 (simox) 2.1E18 1250 24 2.0 yes
S2 (control) yes
4 S6 (simox) 1.8E18 1250 16 1.8 yes
(31854) S7 (simox) 1.8E18 1150 3 0.8 yes
S12 (simox) 2.0E18 1250 16 2.5 yes
S16 (simox) 2.2E18 1250 2 1.9 yes
S19 (simox) 1.8E18 1250 2 1.7 yes
Table 7.2. Defect parameters in the nepilayer of SOI devices.
Et [eV] Nt/ND an [cm2] Possible batch batch batch batch
ap Origin #1 #2 #3 #4
Ec 0.20 1.0E1 1.28E18 V30 complex *
EC 0.30 1.2E2 1.58E18 V20 complex *
Ec 0.12 .6.2E2 5.74E19 Mg, Fe *
Ec 0.65 1.4E2 2.76E11 Ge, Be *
Ev + 0.13 7.2E2 2.1E18 *
Ev + 0.19 3.7E2 1.81E18 Cu, Mn *
Table 7.3. Defect parameters in the pwell of SOI devices.
Et [eV] Nt/ND Op [cm2] Possible batch batch batch batch
Origin #1 #2 #3 #4
Ev + 0.45 2.91E17 Ag, Fe *
Ev + 0.50 1.29E17 Fe,Se, Ge *
Ev + 0.55 3.94E16 Cd, Mn, Cu *
Ev + 0.65 7.75E2 5.96E16 Fe, Cd, 0 *
Table 7.4. Summary of the results for batch2 devices.
Slice Diode n Leakage VB CV Electron Hole
No. [A]at 5 V [V] trap Nt/ND trap Nt/NA
S19 p+/n 1.01 3E12 35 normal 2.6E3 none
control n+/p 1.08 3E12 19 normal 3.3E2 D:3.77E2
S1 p/n 1.15 9E12 34 step 2.35E3 none
SIMOX n+/p 1.39 2E10 14 normal 3.16E2 A:l.5E2
B:2.5E2
C:3.6E2
S4 p+/n 1.16 1E10 35 step 2.86E3 none
n+/p 1.39 2E10 13 normal 3.1E2 A:1.5E2
B:2.5E2
C:3.0E2
Table 7.5. Summary of the results for batch4 devices.
Slice Diode n Leakage VB CV Electron Hole
No. [A]at 5 V [V] trap Nt/ND trap Nt/NA
S2 p+/n 1.05 8.5E12 38 normal  
n+/p 1.27 5.3E10 14.5 normal M:4.3E2 C:7.75E2
S6 p+/n 2.52 2.6E7 44 abrupt J:1.0E1 
n+/p 2.05 1.0E7 9.3 abrupt  
S7 p+/n 1.74 1.9E9 57 abrupt 7.1E2 
n+/p 1.99 1.6E9 9.5 abrupt  
S12 p+/n 1.52 2.4E9 86 step 1.19E1
n+/p 1.52 2.6E8 10.5 normal 
S16 p+/n 1.54 2.1E9 68.2 step K:1.2E2 L:7.2E2
n+/p 1.54 8.2E9 10.6 normal  
S19 p+/n 1.54 1.9E9 72 step < 1.4E 2 
n+/p 1.61 1.3E8 11.5 normal i 3.8E3 3.03E3
Table 7.6 Electron and hole traps observed in batch 4 devices.
Trap Et [eV] a [cm2] Observed in
C Ev + 0.65 5.96E16 pwell
J Ec 0.20 1.28E18 nepilayer
K Ec 0.28 1.58E18 nepilayer
L Ev + 0.13 2.00E18 nepilayer
M Ec 0.53 7.12E17 pwell
Table 7.7 List of possible defect origins for the SIMOX based SOI devices.
Trap level Et [eV] Donor/acceptor Possible Origin
A Ev+0.53 D Mn
Ev+0.53 A Cu
Ev+0.54 Te
Ev+0.55 A Cd
B Ev+0.50 D Fe, Se, Sr
C Ev+0.61 D 0, Fe
I Ev+0.40 D Fe, Sn
Ev+0.45 D Be, Ag
E Ec0.11 A Mg, Fe
F Ec0.62 D Ge, Sr
Ec0.67 D Be,Rb
H Ev+0.18 A Cu, Mn
Ev+0.21 A Na
J Ec0.20 D V3O, VO
to 0.22__ Cd,Ti,W
K Ec0.27 D V20
Ec0.30 D Au,Ge
M Ec0.52 D S
Ec0.55 D Fe
* "D" denotes "donorlike" trap (i.e. a = 1015 1017).
"A" denotes "acceptorlike" trap (i.e. a = 1012 1015 cm2).
CHAPTER EIGHT
DEVELOPMENT AND DEMONSTRATION OF AN
ACCURATE FORWARDBIAS CAPACITANCE SPECTROSCOPY
8.1. Introduction
Measurement of the forwardbias capacitance has been suggested as the nondestruc
tive electrical technique for the investigation of interface state in the metalsemiconductor
contact. This technique has been called by various names such as Schottky spectroscopy,
accurate phase capacitance spectroscopy, and differential voltage capacitance spectroscopy
[18,19,20,89,90]. The principal idea is that sweeping the majoritycarrier quasiFermi level
throughout the bandgap enables the modulation of charge states of interface defect lev
els. This can be accomplished by applying the dc forwardbias to the Schottky diode with
the small ac signal superimposed on the dc forwardbias. The charge state modulation of
interface state contributes to the junction capacitance if the applied ac signal is in the low
frequency regime. The theory for the analysis of forwardbias capacitance is welldeveloped
[18,19]. The problem in the application of this theory arises from the difficulty to measure
the capacitance under the large current flow.
Admittance spectroscopy for studying the MOS devices is not directly applicable to
the forwardbias capacitance measurement of a Schottky diode because the conductance
component due to thermionic emission overwhelms the susceptance (wC). A bridge method
using three equal registers was suggested by Barret and Vapaille [91]. This method is te
dious and timeconsuming even if only one data point is being obtained since the bridge
balance is obtained by changing the bias. During this procedure, one can easily overload the
lockin amplifier or currentsensitive preamplifer. Greve modified Barret's method using the
currentsensitive preamplifier and floating ac signal source [92]. His biasing method has the
problem of overloading the currentsensitive preamplifier. The floating signal source easily
pick up extra noise. Wu, Evans and Yang developed the accurate phase capacitance spec
troscopy (APCS) and the differential voltage capacitance spectroscopy (DVCS) [20,89,90].
Both systems still have the weak points. The small error of phase setting causes the large
error in measuring the susceptance in the APCS technique and the real value of dcbias
to the Schottky diode is unclear in the DVCS technique. Also, the possibility of the large
differential dcunbalance input into the differential preamplifier is high. Even if the differen
tial preamplifier is used, the large dcunbalance input into the differential preamplifier still
causes the overload of lockin amplifier and it brings about the decrease of its sensitivity.
The best way to implement the forwardbias capacitance measurement system is to null
out both the inphase component of a bridge circuit and the unbalanced dc differential
voltage imposed on the differential preamplifier at the same time. The circuit design and
implementation of an accurate forwardbias capacitance spectroscopy will be presented next.
8.2. Accurate Forwardbias Capacitance Spectroscopy
The prime concept of the accurate forwardbias capacitance spectroscopy is obtained
from the careful observation of currentvoltage characteristics of a Schottky diode. One
must keep it in mind that the dc resistance, Rda and the ac resistance, Rac of a Schottky
diode are different. Let us discuss this matter more theoretically. One must note that the
currentvoltage characteristics of a Schottky diode can be empirically expressed by
I = Is[exp(qVF/nkT) 1] (8.1)
or
I = Isexp(qVF/nkT) for V > kT/q. (8.2)
The dc bias of a Schottky diode is related with its dc resistance as
Rdc = VF/Idc.
(8.3)
If the imaginary part of the ac impedance of a Schottky diode is neglected, the ac resistance
can be expressed by
0I nkT
nkT RDC
R.C = (5 )  (8.4)
S R for VF > kT/q. (8.5)
q VF
From the above equation, the following relationship can be derived as
Rac < Rdc for VF > kT/q (8.6)
Now one can conclude that even if the Rac is balanced, the Rdc can be unbalanced in the
previous bridge systems attempted by other groups. This kinds of dc unbalanced voltage
input into the differential preamplifier easily drives the lockin amplifier into the overload
condition or nonlinear regime such that the sensitivity of measurement becomes instrumen
tally limited.
An accurate forwardbias capacitance spectroscopy can be accomplished as in Fig.
8.1. A small signal is supplied from the builtin signal generator in the PAR 124A lock
in amplifier and is superimposed on a debias voltage, which is applied by the HP 6112A
dc power supply. A variable resistor is connected in parallel to the Schottky diode. The
resistance value of variable resistor, Rp, will be determined from the maximum conductance
of a Schottky diode. This means that the parallel resistor will have Rac of which value is
close to the inverse of a Schottky conductance at the maximum forwardbias unless the
majority carrier quasiFermi level sweep over the conduction band edge. Since the resistor
has the same value in dc and in ac, the overall resistance value of a Schottky diode plus
the variable resistor is close to the resistance of a variable resistor. Thus, the bridge circuit
will be balanced in the terms of both ac and dc if the another variable resistor, R1, is
adjusted to null out the inphase current. Then, the Schottky diode will be forwardbiased
by one half of the applied voltage. The sensitivity of a lockin amplifier can be set at
its maximum position. From low to high bias, one can do fine adjustments continuously
Schematic diagram of the accurate forwardbias capacitance spectroscopy.
Figure 8.1.
measuring as many data points as one wants without overloading the lockin amplifier
because the adjusted value, R1, of the potentiometer is close to the value of Rp. Thus,
only a small adjustment is necessary at each different bias step. This means that the AFCS
measurement does not necessiate the wide range of potentiometer. This greatly improves
the efficiency of measurement.
The voltage output of the lockin amplifier can be expressed by
V8A
Vo = [(1/R G)coswt + wCsinwt] (8.7)
where A is the voltage gain of the lockin amplifier and Va is the magnitude of a small signal.
When the phase is set to 900 with a small error of Awt, which arises from the instrumental
limit of a lockin amplifier, after the inphase component is nulled out, the output becomes
VaA
Vo(Awt) = [(l/R G)Awt + wC] (8.8)
and the measurement error of wC is
A(wC) = (G 1/R)Awt. (8.9)
If the conductance component is nulled out, the error will be very small even in the presence
of a significant Awt.
The measurement procedure can be summarized as follows:
1. The barrier height of a Schottky diode must be decided from the CV and the IV
measurements. Make sure the applicalbe maximum forwardbias from the barrier
height.
2. The turnon voltage must be decided from the IV characteristics of a Schottky diode.
3. The conductance at the maximum forwardbias voltage must be determined and the
value of a parallel resistor, Rp, to the schottky diode can be determined in the range
of Rdc > Rp V Rac.
4. The measurements of the forwardbias capacitance can be carried out up to the max
imum forwardbias which is equal to q(B.
As one expects from the measurement procedures described above, we can do the very
fine adjustments of a bride circuit using the precision potentiometer.
8.3. Results and Discussion
The accurate forwardbias capacitance spectroscopy was applied to the Al/ntype GaAs
Schottky diode fabricated by the metalorganic chemical vapor deposition (MOCVD) tech
nique. The measurements were carried out at room temperature with changing the small
signal frequency from 2 HIz to 500 Hz. Figure. 8.2. shows the result of an accurate forward
bias capacitance spectroscopy measurement. It clearly shows the variation of forwardbias
capacitance, Cis versus frequency in the lowfrequency regime. The forwardbias capaci
tance, Ciso makes a peak at the bias of 0.65 V corresponding to the descrete interface state
located at the Ec 0.22 eV because the barrier height of Al/nGaAs is 0.87 eV. To know
whether this level is a real descrete interface state or a demarcation level, further investi
gation is needed. The relationships such as cutoff frequency versus temperature and time
constant of carrier exchange mechanism can be used for this purpose. The detailed steps
for this purpose are well explained in the reference [18,19]. If it is assumed that the Ec
0.22 eV is a real interface state defect, the estimation of the physical parameters such
as defect density, time constant of overall carrier exchange, and cutoff frequency can be
executed as follows. One can deduce the interface state capacitance, Ciso using the following
relationship
Ciso
is = 1+ w2r2 (8.10)
First, using Eq. (8.9), one can estimate the time constant of carrier exchange mechanism,
r because one knows the forwardbias capacitances, Ci,(at 2 Hz) and Cis(at 5 lIz). The
time constant, r, will be 5.3 x 102 second at room temperature. The cutoff frequency is
 !
I
I
*
I
I
 I
I
S 1
I
I
 I
/
I
/ U
!
. *
2 Hz i*O.
I
I
0
f
!
5 Hz,
0*
e"
0
S,
S
20 Hz
S...o.... ** *** .. 50 Hz
*I. l Je *o * **o e. e t I I
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2
VF (V)
Figure 8.2. Forwardbias capacitance of the Al/nGaAs Schottky diode of a small
area versus forwardbias with varying the smallsignal frequency.
12001 1 1 I 1 I 1 1" i 1 1 i1 
defined as
fe r. (8.11)
The estimated cutoff frequency, fc, is exactly 3 Hz. The peak value of Ciso is estimated
as 1625 nF. The interface state density can be easily deduced from Eq. (2.32). For the
convenience, this equation is rewritten as
Ciso = (1 F). (8.12)
Since the Ciso makes a peak at F = 0.5, the estimated interface state density, Nis, is 1.06
x 1012 cm2eV1. This value is very reasonable. Furthermore, considering both the time
constant, r, and the Ec 0.22 eV, this interface state level can be concluded not as a
demarcation level but as a real descrete interface state defect because the demarcation level
is located far below the Ec 0.30 eV [18,19]. The demarcation level, Edn, is expressed by
E BCpNv
Edn = I)Bp kTln(CN) (8.13)
CNNc
where the energy Edn means that the demarcation level is located below the conduction
band; and 4Bp means the barrier height for hole carriers and its value is equal to Eg 4Bn,
i.e., 0.55 eV; and Cp (CN) denotes the capture coefficient of holes (electrons). From Eq.
(8.13), one can easily conclude that the demarcation level is located far below the Ec 0.22
eV level because the last term in Eq. (8.13) cannot be physically over 0.33 eV.
In this study, an accurate forwardbias capacitance spectroscopy (AFCS) was used to
measure the interface state capacitances in the Al/ ntype GaAs Schottky diode grown by
the MOCVD technique. It is fully demonstrated that the accurate forwardbias capacitance
spectroscopy is very powerful and flexible for the study of interface states in the metal
semiconductor system. The AFCS method is more exact and more efficient than the other
techniques such as Schottky capacitance spectroscopy, accurate phase capacitance spec
troscopy, and differential voltage capacitance spectroscopy. The AFCS is very useful for the
nondestructive study of the interface states in the real Schottky diode of a small area.
CHAPTER NINE
CONCLUSIONS AND RECOMMENDATIONS
The major accomplishments of this study are:
(1) A generalized theory for determining the fieldenhanced thermal emission rate by reverse
pulsed deep level transient spectroscopy (RDLTS) is developed. The developed theory was
applied to measure the field enhancement at the real high electric field up to 7 x 105 V/cm.
The previous published data were limited to less than 105 V/cm. This method enables us
to measure the field enhancement at the real high electric field just before the junction
breakdown voltage.
(2) A determination method for temperaturedependent capture cross section is developed.
This method is simple and direct. The previous methods are very complicated and undirect
because a lot of approximations are used in the procedure of theory development.
(3) An accurate determination method for DX center and shallow center densities is de
veloped using thermally stimulated capacitance (TSCAP) technique. The free electrons on
the conduction bands of AlxGalxAs (0.3 < x < 0.4) are supplied comparably by both the
shallow center and DX center at room temperature.
(4) Proton, electron, and neutron irradiation produce the different features of the deep level
defects in AlGaAs. The highenergy proton and neutron irradiation rendered the similar
damage to the lattice, whereas the lowenergy proton and electron produced the same deep
level defects. The Ec 0.99 eV level assigned as the defect clusters, which had a big capture
cross section, was observed in the neutron and highenergy proton irradiated AlGaAs. The
Ec 0.55 eV level ascribed as the Bevacancy complex was observed in the lowenergy
proton and electron irradiated AlGaAs.
(5) The EL2like group (Ec 0.76 eV and Ec 0.86 eV) were observed in the Tedoped
Alo.3sGao.62As irradiated by 300 keV and 1 MeV proton and 1 MeV electron. The Ec
0.86 eV level was also observed in the preirradiated specimen but its density was increased
significantly, by the order, after irradiation. The Ec 0.86 eV level is assigned as AsGa and
the Ec 0.76 eV level is assigned as ASGaVAs.
(6) An accurate forwardbias capacitance spectroscopy (AFCS) is developed for the Schottky
interface state study. This method is really applicable to the real Schottky diode of small
area. This approach is a powerful method for the nondestructive characterization of the
metalsemiconductor contacts in the VLSI integrated circuits.
The following topics are suggested as useful future efforts based on the developed methods
and approaches of the present study:
(1) A generalized theory for determining the fieldenhanced thermal emission rate by RDLTS
must be applied to the GaAs/AlGaAs materials and devices characterization to decide on
the potential well type of the deep level defects. The flexibility of this theory will be limited
by the capabilities of DLTS system such as the applicable short pulse width and sensitivity
because the fast part of capturing transient capacitance is made use of in this theory. The
computerized DLTS system with the fast circuit option in connection with the excellent
pulse generator is recommended for this study.
(2) More fundamental study about the DX center is necessary in connection with the new DX
model that the DX center is not the donor + vacancy complex but the substitutional donor
itself. The DX center is related to the multiconduction band structure of AlGaAs. From
this study, the DX center has the polarization potential well. Based on the assumption that
the substitutional donor can communicate with the direct conduction band (Fconduction
band) and the indirect conduction bands at the same time, it is worthwhile to investigate
whether the polarization well can be assigned to the DX center.
(3) A determination method for the temperaturedependent capture cross section must be
applied to various deep level defects in the GaAs and AlGaAs epitaxial layers. The previous
