Energetic Distribution of Interface States -Between Metal and V...
Transcript of Energetic Distribution of Interface States -Between Metal and V...
VI Energetic Distribution of Interface States -Between Metal and 111-V Semiconductor
6.1 Introduction
Metal-semiconductors contacts have been widely investigated over the years because of
two reasons, first the fundamental understanding of their physical properties and second·
their technological application in devices. The properties. of interface states between metal
.. and semiconductor depend on the growth technique, associated processing parameters and
surface states. The process used for depositing metallic layers on a semiconductor strongly
affects the properties of Schottky co"ntact because the process itself can create defects in
the interface region of metal and semiconductor. Some of these phenomena give rise to.
interface states which affect the barrier height and transport processes through the Schottky
contact. The performance and reliability of Schottky diode and Schottky barrier based
devices, such as metal semiconductor field effect transistor(MESFET), detector and high
electron mobility transistor (HEMT) are affected by the interface states between metal and
semiconductor. Interface states between gate metal and semiconductor reduce the device
gain and increase the threshold voltage and gate leakage current, which can cause high
power consumption and noise in circuits[I,2]. It has been shown[3,4] that an important
source of low frequency noise in field effect transistors is the interaction of electrons in the
channel with interface states. Jantch[5] has shown that the origin of l/f noise in field effect
transistors is due to the random walk of mobile carriers in the interface region. The low
frequency noise behavior is important for several applications of HEMTs. AIGaAs/InGaAs
based pseudomorphic HEMTs are enjoying significant success in microwave and millimeter
102
wave power applications[6,7] and are the fastest commercially available 3-terminal devices
to date. There is strong interest in the development of efficient switches in the power ranges
between 100 kW-1 MW and well above 1 MW. In the latter category, the applications include
improved control over power distribution on the electricity grid, and electrical sub-systems
in electric automobiles, advanced aircraft and combat vehicles. It is anticipated that the
packaged semiconductor switches will need in these applications at temperatures in excess of
3000 e without liquid cooling. CaN based high power Schottky rectifier and MESFET will
be ideal candidate for these applications. Interface states between GaN and metal electrode
in high power Schootky diode is detrimental for switching speed and leakage current. Very
few investigations have been published so far regarding this interface defect density and.
their energetic distribution in CaAs based sub-micron devices and there is no report on the
investigation of interface states in CaN based devices. One reason might be the difficulties in
experimental investigations and their theoretical interpretation. Recently, Balakrishnan et
aI., have shown[8] the effect of interface states under passivated source-gate and gate-drain
regions on the transconductance dispersion in GaAs MESFETs.
Starting from the classical work[9] by Bardeen in 1947, several investigations have been
attempted to clarify the properties and origin of interface states[10]. Though the unified
defect model of Spicer and co-workers[1l,12] can explain the mechanism of Schottky barrier
formation on 111-V semiconductors, so far no agreement on the basic properties, such as
whether these states are of a continuous or of a discrete nature, their energetic location and
their electrical parameters have been achieved. In practical cases, Schottky contacts are
not intimate metal-semiconductor contacts, but separated by a thin layer. This inadvertent
interfacial layer can arise from processing steps and due to various reasons such as cross
diffusion, out-diffusion and chemical reaction between metal and semiconductor. Moreover,
111-V semiconductors used in these devices are plagued with comparatively high density of
surface states compared to elemental semiconductors and the physics of Schottky contacts
on 111-V semiconductors is complicated because of the fact that the Schottky barrier height
103
of III-V semiconductors such as GaAs is weakly dependent or independent of metal work
function due to surface Fermi level pinning[10-12]. Hence, the current transport through
Schottky contacts in III-V semiconductor based devices is affected by the presence of thin
interfacial layer and defect states at the interface between metal and semiconductor.
There are several methods [10,13] to estimate the interface states density in metal
semiconductor and metal-insulator-semiconductor structures. These methods are mainly
based on high and low frequency capacitance and conductance or sub-threshold current
measurements. All these methods depend on the precise determination of Schottky gate
capacitance, including the sub-threshold current technique which uses depletion layer capac
itance during weak inversion. In case of sub-micron gate devices, because of the small gate
area, precise determination of capacitance(which will be around IpF) and detection of any
change in capacitance are not possible due to the limitations ofmeasuriIig instruments. As
the gate length is getting scaled down gradually, it is desirable to avoid any method based on
capacitance measurements. There are several attempts to determine interface states density
from current-voltage(I-V) characteristics[10,14-16]. All these works have used either anequi
libriumapproach or used a priori assumption that interface state distribution is exponential.
Maeda and co-workers[17,18] have proposed a non-equilibrium approach, which is based on
the fact that departure from ideal I-V characteristics of III-V Schottky diode is due to the
change in occupation of interface states with applied bias. Occupation of the interface states
changes due to emission and capture of carriers, which in turn changes the Schottky barrier
height with applied bias.
In this work we have presented a method based on the model proposed by Maeda and
co-workers[17,18] for determining the distribution of interface state density between metal
and semiconductor Schottky contact from I-V characteristics. We have used our techniqU{
to determine the density and the· energetic distribution of interface states in metal-GaA~
(MESFET) and metal-AIGaAs (HEMT) Schottky contacts[19]. The results match well witb
the results obtained by photoemission spectroscopy by other groups. Finally we have used
104
qV. I.
Ec t-t--t-------y-----t- E
FS
qV
E 9
Figure 1: Energy band diagram for metal/interfacial layer / n - semiconductor.
this technique to determine interface density of states in metal-GaN Schottky contacts.
6.2 Theoretical Basis
In case of III-V semiconductor Schottky contact, the semiconductor surface is considered
to be covered with an inadvertent thin layer of native oxide. At the boundary between the
semiconductor and oxide layer, high density of interface electronic states is expected in the
forbidden gap[1l,12]. The interface states of the semiconductor are occupied by electrons up
to the Fermi level. Energy band diagram for metal/interfacial layer / n-type semiconductor
is shown in Fig.I.
The parameter <Po is known as the as the neutral level. The shaded area indicates the
occupied interface states. The occupation of the interface states produces a space charge,
which causes a potential difference in the interfacial layer and reduces the effect of work
function on the barrier height. The interfacial layer model[20] of the actual Schottky barrier
successfully explains the dependence of the Schottky barrier height on the metal work func-
105
tion and based on the thermal equilibrium conditions at zero applied bias(in this case Fermi
level is same throughout the sample). Ma~da and co workers[17,18] extended this model in ~.
case of non-equilibrium situation, i.e. when a .bias voltage is applied to the Schottky bar-
rier. When a forward bias is applied, Fermi level in the metal side(EFm ) differs from that
of semiconductor side(EFs ). The Fermi level of the interface states EFi should be between
these two Fermi levels. Space charge due to the interface states will change depending upon
the position of the Fermi level of the interface states EFi and produces a change in Schot
tky barrier height. The shift in EFj due to the space charges from the equilibrium position
will cause non-ideality in the forward I-V characteristics. Electrons on the interface states
communicate with conduction band of semiconductor by capture and emission and with the
metal by tunneling. When a voltage is applied, the occupation of the interface states is
determined by the Fermi level EFj of the interface states as shown in Fig.l. Depending .
on the relative magnitude of three communication processes (capture, emission, tunneling)
the level EFj will be higher than or equal or lower than the semiconductor bulk Fermi level
EFs . It has been assumed that the thickness of the interfacial layer is much smaller than
the depletion layer width. According to interfacial layer model[20] the zero bias Schottky
barrier height <PB is represented by
<PB = ,(<Pm - X) + (1-,) [~g - <Po] (1)
where, = 1/(1 + q8~S). X is the electron affinity of the semiconductor, Ds is the interface
energy state density(eV /cm2), Ej is the dielectric constant of the interfacial layer, <Pm is the
work function of metal, 8 is the thickness of the interfacial layer, <Po is the neutral level of
the interface states measured from the top of the valence band and Eg is the band gap of
the semiconductor. When the bias is applied to the Schottky barrier, the occupation of
electrons in the interface states changes, thus altering the amount of space charge created
due to interface states. This, in turn, changes the Schottky barrier height <PB and can be
given by,
106
<PB = (<Pm - X) + ~: [( ~g - <PB - <Po) Ds + VDSb] (2)
where Dsb is the density of interface states due to applied bias V. The second term in the
square bracket in Eqn.2 represents the density of occupied interface states due to the applied
voltage V and the first term represents the density of electrons in the interface states under
zero bias. The term V Dsb is related to the Fermi level EFi of the interface state above the
metal Fermi level EFm through the relation
VDsb Ds q
(3)
Taking the metal Fermi level EFm as the zero energy reference, the change in the barrier
height /).<PB due to applied bias from Eqn.l and 2 is given by,
/).<PB = (<pB - <PH) = (l_,)EFi q
(4)
The calculation of the change in Schottky barrier height due to incremental change in ap-
plied bias will depend on the trap kinetics of the interface states between metal gate and
semiconductor. As already discussed, on the application of the voltage 'bias, the occupation
of the interface states changes through the capture, emission and tunneling processes. The
electron emission probability e by the interface states can be given by
e = eo exp ( - q<p:B; E) (5)
where eo is a constant and depend on the electron capture cross-section of interface states
and effective density of states of the conduction band, E is the energy of the electron in
the interface states, KB is the Boltzman constant and T is the temperature. The electron
capture probability c of the interface states from the conduction band can be given by
_ 0 (_ q(<PB - V)) c - c exp kBT (6)
107
where Co is a constant and depends on electron capture cross-section and electron concen-
tration in the conduction band. The occupation probability f of the interface states can be
given by
f = c e+c+t
(7)
where t is the tunneling probability of electrons from interface states to metal. In case of
zero forward bias, the Fermi levels in metal and semiconductor are aligned. Now, let us
suppose the applied forward bias is being increased in steps, of .~ V. After the first step, the
Schottky barrier changes from <PB (zero bias barrier height) to <p1 and the change in barrier
height becomes ~<p1. Due to change in forward. bias, Fermi level at the interface EFi also
changes by an amount ~EFi due to the change in the occupation of the interface states
from the equilibrium situation. The modified barrier height ( <p1) and the new position of the
Fermi level(E}J will be
(8)
After N increment of the forward bias of equal step, barrier height and the position of the
Fermi level can be given by
",N ",N-l "",N EN EN- 1 "EN 'l'B = 'l'B + ll'l'B' Fi = Fi + II Fi
Eqn.(l) can be expressed as
and after application of a forward bias step ~ V, Eqn.lO becomes
From Eqn.8, 10 and 11, we can get
108
(9)
(10)
(11)
al ~E}i q(1 + al) ~V
where al = q8DsI/f.i and similarly after Nth step, we can get
aN ~E~ q(1 + aN) ~V
(12)
(13)
Now, the variation of ~EFi with applied forward bias has to be determined. In the steady
state condition, if we put f = 1/2, we will have c = e + t. After application of first bias
step ~ V, the capture and emission probabilities can be given by
(14)
We can get, after assuming ~ V to be very small and neglecting higher order terms in
Eqn.12,13 and 14
1
Similarly after Nth step we can get
·N ~EFi 1 ~V - 1- t
cN(l+C>N}
(15)
(16)
The I - V characteristics of a rectifying metal semiconductor contact is normally described
by the thermoionic relation
(17)
where V is the applied bias, T the temperature, n the ideality factor, kB the Boltzman's
constant and q is the electronic charge. Is is the saturation current and related to the metal
semiconductor barrier height through the relation
(18)
109
where A is the area of the diode, A ** is the effective Richardson constant and <PB is the
Schottky barrier height. From Eqn.18, we can get
(19)
Now using Eqn.12 and 15, we can express Eqn.19 as
kBTlnJl-lnIO =_ al +1
q ~V (1 + al) (1- Cdl~al») (20)
The reciprocal of the term on the left hand side in Eqn.20 is the ideality factor n, which can
be denoted asnlafter the application of first forward bias step of ~ V. The term al can be
extracted from the·· above relation as
(21)
Finally, the density of the interface states can be expressed as
(22)
and similarly
DsN = ~ (1 - _t ) (nN - 1) qo eN (23)
The distribution of interface density of states can be given by
Ds = ~ ;~ (1 - :j) (nj - 1) J
(24)
It is clear from Eqn.24 that density of interface states is proportional to the ideality factor
n. One can now proceed to determine the energy distribution of interface states from the
bottom of the conduction band Ec through the relation
110
q¢>B - EFi
q(¢>~ + ~¢>k + ... + ~¢>~) - (E~i + ~E~i + ... + ~Ef.i) o (a1) 1 ( aN) N q¢>B + ~EF· + ... + ~EF· 1 + al l 1 + aN l
q"-O + ~ ( aj ) ~Ej . 'l-'B ~ 1 + a. F,
j J
6.3 Experimental Details
(25)
(26)
The devices used in this work are GaAs MESFET and GaAIAs/InGaAs pseudomorphic
HEMTs(pHEMTs). MESFETs were fabricated by ion-implantation. Ti/Pt/ Au was used as
the gate metal. The gate dimensions of the MESFET are Lg =0.8JLm, z=550JLm. More details
of the MESFET structure is given in Ref.8. The pHEMT structure"was grown by molecular
beam epitaxy. The layer structure for pHEMT features, from bottom to top, a 1JLm thick
undoped GaAs buffer, 130A undoped Ino.2Gao.8As channel, 30A undoped Alo.23Gao.77As
spacer, 600A Si-doped(5xlO12cm-2) Alo.23Gao.77As supply and 400A n+(5x1018cm-3) GaAs
cap layer. The device was grown on a semi-insulating GaAs(100) wafer. The epitaxial layer
was grown at 500°C. The V /111 flux ratio was maintained in the range of 30 to 50 for As-
stabilized condition. Following .epitaxial growth, the active device was isolated by a mesa
. etch and Au:Ge/Ni/ Au ohmic contact was deposited and annealed at 400°C for lOs by rapid
thermal processing. A gate recess etch was then performed to allow the gate metal to set
directly on the AIGaAs electron supply layer. Au/Ti/ Au Schottky gate was formed by e
beam lithography. The gate dimensions of pHEMT are Lg =0.2JLm, z=200JLm. The forward
and reverse I-V characteristics at various temperatures were measured using the Keithley 428
current amplifier, 230 voltage source and liquid nitrogen cryostat. MESFET was fabricated
at SPL, Delhi and pHEMT at NEe, Japan.
In this work, we have investigated the I-V characteristics of two Au/Pt n GaN Schottky
contacts from Ref.21. The GaN layer was grown on Ah03 substrates by radio frequency
111
plasma-assisted MBE. In one set of devices, Schottky contacts were formed on one sample
after a conventional surface cleaning. The conventional cleaning process involves sequential
. rinsing in acetone, isopropyl alcohol and de-ionized water prior to lithography for defining
the contact area. After lithography, the samples were rinsed for 60 s in Hel and 30 s in
buffered HF to remove the native oxide. In case of the second set of devices, samples were
first cleaned in the conventional way then boiled in (NH4h S solution for 20 min before
making Schottky contact. The Pt (400 It)/ Au (1500 It) Schottky contacts were fabricated
by electron beam deposition.
6.4 Experimental Results and Discussions
6.4.1 GaAs-MESFET and GaAIAs/InGaAs-pHEMT
The forward I-V characteristics of MESFET and pHEMT at room temperature are shown
in Fig.2. The figure also shows the fitting with the thermionic emission mechanism. The
excess current in low bias region with high ideality factor suggests a deviation from the
thermionic emission mechanism. At larger bias, the thermionic emission dominates and the
current increases exponentially with bias according to Eqn.18. The temperature dependence
of the saturation current has been used to obtain the Schottky barrier height. The inclusion
of the ideality factor n in the saturation current is necessary to determine the correct Schot
tky barrier height ¢B from the extrapolated saturation current Is . It has been reported by
Ashok et al[22] and Shewchun et al[23,24] that inclusion of the ideality factor in Is is neces
sary to explain the experimental data(Is, ¢B and A**) for Au-GaAs and Ni-GaAs Schottky
diodes and Si MIS solar cells respectively. Fig.3 shows the linear dependence of Is/T2 on
lOOO/nT for MESFET and pHEMT, the slope being proportional to barrier height. In Fig.3,
both In(Is/T2) vs. lOOO/T and In(Is/T2) vs. lOOO/nT are plotted for comparison and the
deviation from linearity at low temperature for 1000/T plot is evident. The ideality factor
at OAV bias has been used, where thermionic emission is dominant. The barrier height ¢B
for the Ti/Pt/ Au-GaAs Schottky diode in the MESFET is O.gleV which agrees well with
112
10-6 10.6
T = 300K T = 300K 10.7
10-8
- 10-8 <C --c CI)
10.10 10.9 ~ ~
:l (.)
10.12
(a) 10.11 (b)
o 0.1 0.2 0.3 0.4 0.5 o 0.1 0.2 0.3 0.4 0.5
Forward Voltage (V) Forward Voltage(V)
Figure 2: The forward J- V characteristics of the Schottky gate of (a) MESFET and (b) pHEMT. Solid lines are the simulated data for thermionic emission and symbols are measured data.
the reported values in case of GaAs Schottky diodes with a very thin interfaciallayer[22].
¢>B for the Au/Ti/ Au-AIGaAs Schottky diode in pHEMT is1.16eV which agrees well with
the results reported by Best[25]. High barrier manifests the presence of a thin interfacial
layer between gate metal-GaAs and metal-AIGaAs in MESFET and pHEMT respectively.
It is clear from Fig.2 that the I-V characteristics at low bias exhibit an excess current,
which cannot be accounted for only by thermionic emission. The origin ofthisexcess current
and the high value of ideality factor in this region suggest recombination in the depletion
region and tunneling as probable causes. We have tried to fit the forward current in the low
biaS region{0-0.3V) with a simulated current comprised of three components, current due
to thermionic emission, current due to field emission and recombination current [10,22,26]'
but we failed to get a reasonable fit. The non-conformity between the experimental data
and the theoretical fit suggests that the current conduction mechanism between the metal
gate and GaAs and AIGaAs in MESFET and pHEMT respectively with the thin interfacial
113
-20 c 10001T 0 1000/nT 0 1000/nT C 10001T
-35.0
-40 -(\I
~ « -37.5 --(\I
t::: -60 C/)
:::::.. r::: -40.0
-80 -42.5 (b)
2.5 5.0 7.5 3 5
10001T or 1000/nT (11K) 10001T or 1000/nT(1/K)
Figure 3: Plots for the determination of barrier height of the (aJ Ti/Pt/Au-GaAs Schottky gate in MESFET and (b) Au/Ti/Au-AIGaAs Schottky gate in pHEMT. The value of ideality factor n is taken at 0.4 V forward bias.
layer is more complicated than an intimate metal-semiconductor contact. Card and Rhod
erick[27] tried to explain the current conduction mechanism in silicon Schottky diodes with
a thin interfacial layer and proposed a correction for the saturation current with a factor
proportional to exp(O.26X¥28), where XS is the mean barrier height and 8is the thickness of
the interfacial layer. However, this model cannot be applied to III-V semiconductor-metal
contacts because of high density of surface states.
Ashok an~ co-workers[22] showed that in case of GaAs, xs determined using Roderick's
method gives too Iowa value to be interpreted as the distance from the conduction band
edge of the semiconductor to that of the insulator. It has been suggested[17,18] that we
have to consider the recombination and generation by the interface states and interface state
assisted tunneling in the current transport mechanism of the metal-interfacial layer-III-V
semiconductor system. In addition, a non-equilibrium theory has to be invoked to take into
account the capture and emission of the carriers from interface states to the conduction
14
3.5 T = 300K T = 300K
3.0 2.0 -c 2.5 -10..
0 -u as LL 2.0 1.5 >-!:: as Q) 1.5 :2
1.0 1.0 (b)
0.5 0.5 '-----'-_--'--_-'----1._--'
o 0.1 0.2 0.3 0.4 0.5 0 0.1 0.2 0.3 0.4 0.5
Forward Voltage(V) Forward Voltage(V)
Figure 4: The variation of the ideality factor n with forward bias applied at the gate in case of (aJ MESFET and (b) pHEMT.
band o,f the semiconductor. The ideality factor n(= k;T d:JI») in Eqn.18 determined from
the forward I-V characteristics at 300K is shown in Fig.3 for MESFET and pHEMT. All
I-V data have been corrected for series resistance using Lee's method[28]. It is clear from
FigA, that the ideality factor n increases with bias to almost 3 and then decreases in both
devices. Such voltage dependence of n has not been observed in Si with interface layers and
is very unique for 111-V semiconductors with high density of surface states. This apparently
strange behavior is considered to be due to interface states between the metal and the 111-V
semiconductor. Spicer and co-workers[11,12] have reported energetic distribution of interface
states between metal and GaAs from photo-emission experiments. The strange behavior of
the ideality factor with applied bias is considered to be related with the distribution of
interface states in the semiconductor forbidden gap[11,12,18]. The density of interface states.
can be determined from Eqn.24 using the forward bias dependent ideality factor n as shown
in FigA. The exact values of the parameters such as dielectric constant of the interfacial
layer Ei, thickness of the interfacial layer b, capture probability c and tunneling probability
115
6
C? ,.. 0.8 0 ..... 5 >< ->
(1) N-
E 4 u ....... CI) 0.4 C
3
0.4 0.6 0.8 1.0 0.6 0.8 1.0 1.2
Ec· E (eV)
Figure 5: Plots of energetic distribution of interface states between metal and semiconductor as a function of energy measured from the bottom of the conduction band in case of (a) A-[ESFET and (b) pHEMT. .
t are not known experimentally. We can however estimate the interface density of states by
assuming fi = 4fo[17,18,20], where fois the dielectric constant of free space, 6 = 30A and t= ,
c/2. It is difficult to measure the thickness of the interfacial layer accurately, but it has been
reported[18,22] 6 ~ 20 - 30A in a similar device structure and processing. Fig.5 shows the
. energetic distribution of the interface states between gate metal and GaAs in MESFET and
gate metal and Alo.23Gao.77As in pHEMT respectively using Eqn.26. The interface density
between Schottky metal gate and GaAs in MESFET peaks at O. 78e V. This value matches
well with the energy level measured by photoemission spectroscopy[11,12]. Spicer and co
workers[11,12] have proposed the existence of two energy levels which are present in the band
gap. The level at O.7eV below the conduction band minimum(CBM) is an acceptor type due
to missing As and the other at 0.geV below CBM is a donor type due to missing Ga on
the surface. Similar work in case of AlxGal_xAs is not available in theliterature. However,
interface density peaking at an energy of 1.02e V might also be due to the As vacancy. In
116
Figure 6: The forward 1- V characteristics of Conventionally cleaned (empty circle) and (N H4hS treated (solid circle) Au/Pt/n CaN Schottky contact. Solid lines are the simu-lated data for thermionic emission.
case of metal-Alo.23Gao.77As in pHEMT, the origin of another broad peak at O.86eV is not
known.
6.4.2 Metal-GaN Schottky contacts
The development of reliable, reproducible and thermally stable Ohmic and Schottky con
tacts is one of the most important areas for GaN device technology. High quality rectifying
and Ohmic contacts 'With low specific resistance are needed for better performance of the
devices. Research on both Ohmic and Schottky contacts of GaN are of current interest.
Schottky contact of various elemental metals like Au, Pt, Pd,Ni, Ti, Cr[29-33] and some
metal oxides like indium tin oxide (ITO)[34,35], cadmium tin oxide(CTO)[36] and also some
silicide like PtSi[37] and NiSi[38] have been investigated. There are still large variations
in barrier heights reported by different workers for standard metals on GaN[29-33]. This
nonuniformity can be due to various reasons, the presence of several transport mechanisms,
the presence of defects in the film, the local stoichiometry variations and the presence of an
117
-c -o 5 -(,) co u. ~ -.--;3 Q)
" 0.2 0.4 0.6 0.8 1.0 1.2
Voltage (V)
Figure 7: The variation of the ideality factor n with forward bias in case of conventionally cleaned (solid line) and (NH4hS treated (dashed line) Au/Pt/n GaN Schottky diode.
interfacial layer between the metal and the semiconductor. Thermal annealing and the effec~
tiveness of the surface cleaning prior to the metal deposition can also affect ,the properties of·
the interface. Shen et al.[35] has shown the effects ofthermalannealing on the ITO Schottky
contacts on GaN. The effective Schottky barrier heights were 0.68, 0.88, 0.94 and 0.95 eV
for nonannealed, 400, 500 and 6000 C annealed samples respectively. Liu et al.[38] have also
shown the thermal annealing effect on Ni and NiSi-GaN Schottky contacts. These results
indicate that an increase of barrier height may be attributed to the change of the properties
of the interfacial states at the metal and GaN interface after annealing. Cao et al. [21] have
shown that the Au/Pt Schottky barrier height on both nand p type GaN samples reduces
with an additional treatment in (NH4hS after conventional surface cleaning prior to the
metal deposition. The role of surface treatments in determining the barrier heights has been
investigated by many others.
In Fig.6, forward I-V characteristics of two Schottky diodes are shown. All I-V char
acteristics have been corrected for series resistance using Lee's method[28]. We have got a
118
10
C") 8 ,.. 0 ,.... ><
6 -> (1) -('II I 4 E (.) -(I)
2 C
0 0.3 0.5 0.7
" ",
Ec - E (eV)
, , ,/
, , / , , , , ,
0.9 1.1
Figure 8: Plots of energetic distribution of interface states between metal and semiconductor as a function. of energy measured from the bottom of the conduction band in case of Con-ventionally cleaned (solid line) and (NH4hS treated (dashed line) Au/Ptfn GaN Schottky diode.
series resistance Rs of 13 K!1 in conventionally cleaned sample and 4 K!1 in (NH4hS treated
sample. Clearly, forward current increases as a result of the (NH4hS treatment. Fig.6also
shows the fitting with the thermoionic emission mechanism of current transport through
the Schottky contact(Eqn.18). The experimental data is showing a clear deviation from the
thermoionic emission current at low and high bias region. The voltage dependence of the
ideality factor n is shown in Fig.7. It shows two distinct picks, one at low bias and another
at high bias region for both the samples. The exact value of the parameters such as dielectric
constant of the interfacial layer ti, thickness of the interfacial layer 8, capture probability c
and tunneling probability t are not known experimentally. We have taken ti = 2to·, 8 = 20
A and the ratio t / c = 0.5 to estimate the interface density of states.
Fig.8 shows the energetic distribution of the interface states for both the samples. The
conventionally cleaned sample shows higher level of interface state density than the treated
sample. Two distinct picks are appearing at around 0.45 eV and Ie V, which are same as the
119
energy levels for V Ga and/or V Ga-ON in bulk GaN( discussed in detail in Chapter III). The
mechanism, which is responsible of these native defects(V Ga and V Ga-ON) in bulk n-type -.
GaN, should also be responsible for these defects on n-type GaN surface. After the treatment
with (NH4hS both the pick heights reduced. V Ga and V Ga-ON are acceptor type defects.
Either H or S from (NH4hS can passivate these interface states.
6.5 Conclusion
The departure from an ideal metal-semiconductor contact inSchottky diode and Schottky
diode based devices has been explained quantitatively. Nonideal I-V characteristics and bias
dependent ideality factor have been correlated with the existence of interface states and a
thin interfacial layer between metal and GaAs in MESFET, metal and AIGaAs in pHEMT
and metal and GaN in Gan-Schottky diode. Using a non-equilibrium theoretical model,
energetic distribution and density of interface states have been determined from the bias
depended ideality factor. This me~hod can be successfully applied to study the interface
states in sub-micron devices.
120
References
1. M. M. Ahmed, H. Ahmed andP. H. Ladbroke, J. Vac. Sci. Technol. B 13, 1519 (1995).
2. M. Shur, Physics of Semiconductor Devices, Prentice-Hall Englewood Cliffs, NJ, 1990.
3. D. L. Lile and M. J. Taylor, J. Appl. Phys. 54, 260, 1983.
4. C.Su, H. Rohdin and C. Stolte, IEDM Technical Digest,pp 601-604, 1983.
5. O. Jantch, IEEE Trans. on Electron Devices, ED-34,1l00 (1987).
6. S. W. Chen et al.,5,201 (1995).
7. J. J. Brown etal., IEEE Microwave Guided Wave Lett. 6, 91 (1996).
8. V. R. Balakrishnan, V. Kumar and S. Ghosh, IEEE Trans. Electron Devices, ED-44,
1060 (1997).
9. J. Bardeen, Phys. Rev. 71, 717, (1947).
10. E. H. Rhoderick and R. H. Williams, Metal Insulator Contacts, 2nd Ed, Clarendon,
Oxford, 1988.
11. W. E. Spicer et al., J. Vac. Sci. Technol. 16, 1422 (1979).
12. W. E. Spicer et al., Phys. Rev. Lett. 44, 420 (1980).
13. D. K. Schroeder, Semiconductor material and device characterization, John Wiley & Sons, New York, 1990.
14. H. H. Tseng and C. Y. Wu, Solid State Electron. 30, 383 (1987).
15. J. M. Borrego, R. J. Gutmann and S. Ashok, Solid State Electron, 20, 125 (1977).
16. Z. J. Horvath, J. Appl. Phys. 63, 976 (1988).
17. K. Maeda et al., J. Appl. Phys. 68, 2858 (1990).
18. K. Maeda et al., Appl. Phys. Lett. 62, 2560 (1993).
19. S. Dhar, V. R. Balakrishnan, V. Kumar and S. Ghosh, IEEE Trans. Electron Devices,
ED-47,282 (2000).
20. A. M. Cowley and S. M. Sze, J. Appl. Phys. 36, 3212 (1965).
121
21. X. A. Cao et al., Appl. Phys. Lett. 75, 4130 (1999).
22. S. Ashok, J. M. Borego and R. J. Gutman, Solid State Electron, 22, 621 (1979).
23. J. Shewchun et al., Appl. Phys. Lett. 35, 416 (1979)~
24. J. Shewchun et al., J. Appl. Phys. 50, 3832 (1979).
25. J. S. Best, Appl. Phys. Lett. 34, 522 (1979).
26.M. Wittmer, Phys. Rev. B 43, 4385 (1991).
27. H. C. Card and Rhoderick, ,J. Phys. D, 4, 1589 (1971).
28. T. C. Lee, C. D. Beting and H. L. Au, J. Appl. Phys. 72, 4739 (1992)
29. A. T. Ping et al., Electron Lett. 32, 68 (1996).
30. S. C. Binari et al., Electron Lett. 32, 909 (1994).
31. J. D. Guo et al., Appl.Phys. Lett. 67, 2657(1998).
32. J. D. Guo et al., J. Appl. Phys. 80, 1623 (1996).
33. A.C. Schmitz et al., J. Elctron. Mat. 27, 255 (1998).
34. W. Gao et al., Appl. Phys. Lett. 65, 1930 (1994).
35. J. K. Shen et al., Appl. Phys. Lett. 72, 3317 (1998).
36. Q. Z; Liu et al. Appl. Phys. Lett. 70, 1275 (1997).
37. Q. Z. Liu et al., J. Appl. Phys. 84, 881 (1998).
122