7. Conclusion
In order to investigate the use of optically transparent indium tin oxide (ITO), a range of
technological steps were developed and optimised during the course of this study. These
include reactive r.f. sputtering, reactive ion etching, careful optimisation for producing ideal
Schottky contacts, annealing for producing good ohmic contacts using ITO and metal ohmic
contacts on ITO itself. A number of novel optoelectronic devices were fabricated as a result
including: transparent metal contact Schottky photo diode
[228], transparent gate HEMT (TG-HEMT) [79],
transparent electrode VCSELs [88] and LEDs and
transparent emitter contact HPTs [2]
.
7.1 ITO Deposition
A reactive r.f. sputtering technique using an argon/oxygen mixture was utilized for the deposition of the Indium Tin Oxide films from a hot pressed target. A number of deposition parameters and their effects on the films were studied in detail. In particular the partial pressure of gaseous oxygen, PO2, in the plasma and the r.f. power, directly affecting the deposition rate, were found to be amongst the most influential. While the conductivity was found to be inversely proportional to the exponential of the PO2, the transparency reduced significantly in the absence of sufficient oxygen. The r.f. power was directly proportional to the film conductivity. These parameters were then optimised to maintain the desired combination of high transparency and conductivity properties of the films and to minimise sputter damage to the substrate material constituting the active device layers. The critical values for PO2 and the r.f. power used were 14e-3 (at a total pressure of approximately 10 mTorr) and 150W respectively.
Using this technique, highly conductive uniform ITO films were obtained. The as deposited sheet resistance, Rsh, was typically 340 W/ which reduced to 27 W/ after annealing. These values correspond to resistivities of 190e-4 Wcm and 15e-4 Wcm respectively and compare favourably to those in the literature. Above 90% transmittance was achieved for these ITO films over the 500nm to 2000nm wavelength range. This corresponds to 2.48eV to 0.62eV energy range which covers majority of the III-V compound semiconductors such as AlAs, AlGaAs, GaAs, InP and InGaAs as well as Si and Ge. For a typical ITO film, the value of the bandgap calculated from the transmission spectra was found to 3.8eV, which is in close agreement with previously published work.
Driven by the need to produce near micron structures with ITO films, a number of
experiments were carried out to develop a suitable method for micro-pattering these films.
Due to the energetic species inside the sputtering chamber, resulting in baking of the
photoresist, liftoff lithography was rendered unsuitable. Wet chemical etching involving a
number of chemicals, such as HCl, HNO3 etc., in various strengths as
well as Zn dust as a catalyst were tried and were also found to be unsatisfactory for this delicate application. Finally a selective dry etching technique based on reactive ion etching (RIE) with
Ar gas was developed; a 1000 thick layer of Al metal was used as the mask. As well as having
a very good selective etch ratio of greater than 4:1 for Al:ITO, the use of Ar as the plasma gas is
both cheaper and safer. Two micron fingers were successfully produced using this etching
technique. The optimum r.f. power and the chamber pressure RIE system was found to be
100W and 11mtorr respectively.
Two distinct sets of post deposition annealing techniques were developed. One of these was
to obtain an ITO/n-GaAs Schottky diode with rectifying contact with low series resistance
while retaining the diode parameters for a good device. A two step process extended over
several hours was used to ensure the reduction of ITO film resistivity and the recovery of
sputter damage from the substrate without affecting the rectifying properties of the device. A
five hour anneal at 240C followed by another five hours at 340C in a N2
ambient was found to produce these desired results.
The other annealing technique was developed to produce a heterojunction photo transistor
with essentially an ohmic contact to the emitter cap layer. Given that the stringent
requirement not to destroy a rectifying junction is no longer applicable, the realisation of an
ITO ohmic contact proved to be more difficult than at first imagined. Traditional annealing
techniques with RTA or dual alloying methods were unable to produce good uniform ohmic
contacts. Thus a new technique involving a step-like annealing method was employed.
Samples were at first gently heated at lower temperatures (240C) for relatively longer periods
(3 minutes) and gradually the annealing temperature was increased to up to 500C where it
was placed for a shorter time (1 minute); the sample was cooled using the reverse
temperature/time combination.
Using both these techniques, the resultant ITO resistivity was found to be approximately
14e-4 Wcm. This compares well with the published data.
Bearing in mind the need to form large metal pads for connection to external circuits, a set of
experiments were performed to assess various metalisation schemes for making contacts to the
ITO films themselves. These metalisation systems were Ni/AuGe/Ni/Au, Ni/Al and
Ni/AuGe/Ni/Ti/Au; as well as the ohmic properties, mechanical adhesion was also a factor in
these assessments. It was concluded that in the absence of an adhesion promoting layer such
as Ni, the use of say only a layer of Au was unsuitable. Comparing the electrical properties of
the three metalisation systems investigated, it is noticeable that their unalloyed characteristics
were very similar (Rsk and rc in the region of 50W/ and 3e-4 Wcm2 respectively) despite
the fact that the sheet resistances of the corresponding ITO material varies over a much
greater range (1273 W/ to 665 W/).
The Rsk after alloying for all three metalisations lie approximately in the range 1 to 2
W/ while rc lie in the range
3 to 8 e-5 Wcm2. Thus all three systems are equally suited for
contacting ITO films as there is very little difference between them.
Following thorough cleaning and pre-metalisation chemical etch to the surface of MBE (Si)
grown, n-GaAs wafers with ND = 1.4e16 cm-3, 1.75
mm thick active layers, near ideal Schottky diodes were
produced using Au and Al as the metal contacts. The ideality factor, n, the barrier height,
fbo, and the
modified Richardson constant, A**, for the Au/n-GaAs devices were 1.04, 0.84 eV and
12.5e6 Am-2K-2; the corresponding values for the Al/n-GaAs diodes were 1.08, 0.77 eV
and 16.0e4 Am-2K-2. These values for both sets of devices are in good agreement with the
findings of other researchers. The Schottky diodes were also characterised over a wide
temperature range to provide a thorough understanding of their d.c. performance. The
results analysed using a simulation model involving the influence of various current
transport mechanisms at different temperatures. Excellent agreement was found between
the measured and the simulated data.
In the quest for replicating near ideal Schottky diodes with an ITO film replacing the
conventional metal contact in the previous set of diodes, numerous experiments were
performed to finally achieve the desired goal. As well as thoroughly cleaning the wafer prior
to ITO deposition, a number of sputtering parameters had to be optimised. Some of these,
such as r.f. power and PO2 and the relevant post-deposition annealing
techniques, have already been discussed earlier. In addition, a thin layer of Indium metal,
100, deposited between the ITO and the semiconductor was found to be effective in shielding
the substrate from sputter damage. Thus ITO/n-GaAs Schottky photo diodes were realised with
the following characteristics: n = 1.1, f
bo = 0.76 eV, dark current at -2V = 50 nA, series resistance = 25.8
W, responsivity to 780 nm radiation = 0.57 A/W and an
external quantum efficiency, h = 91.5%. These are amongst
some of the best results obtained thus far and compare better than many reported in the
literature for similar devices. Thus near ideal, ITO/n-GaAs Schottky diodes with low leakage
current, low series resistance and high responsivity have been fabricated on n-GaAs epitaxial
layers.
The effect of mesa etching, r.f. power, use of a thin layer of indium and gradual annealing at
340C have been studied and optimised to produce near ideal photo diodes. Mesa etching
significantly reduces the surface leakage current thus suppressing the dark current. Increasing
the r.f. power reduces the sheet resistance of the ITO film but causes greater damage to the
substrate; this damage can be minimised by depositing a thin layer of indium on the GaAs
substrate prior to sputtering and the damage can be removed by gradual annealing.
A detailed C-V analysis was carried out to separate the effects of sputter damage from
associated series resistance. It was concluded that sputtering causes two types of damage:
These novel ITO/n-GaAs devices were then subject to cryogenic measurements similar to their
conventional opaque counterparts. From the measurements, the modified Richardson
constant, A**, for these devices were found to be 1.3e4 Am-2K-2, and to the best of our
knowledge this is the first time it is being reported for ITO/n-GaAs devices produced by
reactive r.f. sputtering.
Having thus obtained relatively large ITO/n-GaAs Schottky photo diodes, high speed diodes
can be achieved by shrinking the active region; parasitic capacitances and leakage currents can
be reduced by a combination of proton bombardment and mesa etch isolation techniques.
As an extension of the work with ITO/n-GaAs Schottky diodes, a set of novel pseudomorphic
HEMTs with a transparent gate (TG-HEMT) were fabricated using transparent indium tin
oxide for the first time. These AlGaAs/InGaAs/GaAs pseudomorphic HEMTs (pHEMTs)
with ITO gate electrodes have an optical responsivity greater than 5A/W corresponding to an
external quantum efficiency of greater than 800% for an incident radiation of
l= 780nm. A
set of conventional devices were also fabricated for comparison using opaque Ti Schottky
gates. This was performed in collaboration with another researcher in the group to
demonstrate the viability of such devices and greater detail is available in
[229]
.
The next part of this thesis involved the fabrication of good ITO ohmic contacts to highly
conductive n and p type layers (ND = 5e18 cm-3,
NA = 5e19 cm-3) of GaAs and n-type InGaAs layers
(ND = 3e18 cm-3). The objective was to realise good transparent
contacts to the emitter of HBTs for vertical illumination of an optical signal. The TLM
method was used for assessing the quality of these contacts. In each of these cases, the
transparent ohmic contacts were compared to ohmic contacts made from conventional
metals on the same substrates to enable a direct comparison to be made.
For the ITO/n+-GaAs contacts, the insertion of a thin layer of In metal between the
semiconductor and the transparent film was found to be very effective in enhancing the ohmic
properties of the contact. Although initially this reduced the transparency to nearly 60%, this
parameter was restored to above 90% following the ohmic contact annealing scheme
discussed earlier. It is believed that In acts as a doping species in n-type GaAs thereby
reducing the specific contact resistance. The best transfer resistance, Rt, and the specific
contact resistances, rc, achieved in each case were as follows:
Rt = 0.68 Wmm, rc = 1.2e-4
Wcm2 (n+-GaAs); Rt = 0.49Wmm,
rc = 7.6e-5 Wcm2 (In/n+-GaAs).
Thus the superior quality of the ITO/In/n+-GaAs over the ITO/n+-GaAs ohmic is clearly
demonstrated [230]. The corresponding values for
the conventional contacts realised using the Ni/AuGe/Ni/Au metalisation system, were as
follows: Rt = 0.10Wmm, rc
= 3.1e-5 Wcm2.
The values achieved for these parameters in case of the ITO/p+-GaAs contacts and their
conventional counterparts fabricated using Au/Zn/Au metalisation system were Rt =
0.55Wmm, rc = 3.2e-5
Wcm2 (ITO); and Rt = 0.07Wmm,
rc = 5.1e-6 Wcm2
(conventional) respectively. Although comparably good transparent ohmic contacts were
achieved on p+-GaAs substrates, the increased difference between the transparent and their
opaque counterparts is attributable to the absence of any obvious dopant species in the ITO
contacts. Despite this absence, the main reason behind realising good p-type transparent
ohmic contacts is likely to be due to the high substrate doping alone.
The results for the ITO/n+-InGaAs contacts and their conventional counterparts were as
follows: Rt = 0.37Wmm, rc = 7.1e-5
Wcm2 (ITO); and Rt = 0.10 Wmm,
rc = 3.1e-5 Wcm2 (conventional).
A brief surface morphology examination of the ITO contacts and subsequent comparison with
the conventional metal contacts revealed some interesting results. It was seen that there was
significant "balling-up" on the surface of the conventional contacts to the n+-GaAs substrates.
The diameter of an average "mound" is approximately 5mm -
comparable to, if not larger than, the dimensions involved in emitter finger widths of microwave
HBTs. This effect is reduced but nevertheless it is still present in the p-type contact. By contrast,
the metal on the ITO/n+-GaAs surface was smooth suggesting that there is no liquid phase
reactions involved in the latter ohmic contact and ITO acts like a capping (such as Ti/Au) or
a diffusion barrier (such as ZrB2) layer preventing the intermixing of
the top-most Au layer with the substrate causing the balling effect. Its demonstrated excellent
surface morphology and edge definition is an added advantage in the fabrication of small
geometry optoelectronic devices. This observation is further correlated by the fact that Sn
present in ITO is also known to take part in solid phase reactions to n-GaAs in the formation
of good ohmic contacts. However, like other solid phase schemes, such as Pd/Ge or Pd/Sn,
the resistance is higher than that obtained for the Ni/AuGe/Ni/Au liquid phase n-type
contact.
The penultimate part of the study concerned the study of heterojunction bipolar transistors
fabricated on three different material systems: AlGaAs/GaAs, InGaP/GaAs and InP/InGaAs
respectively [231]. This was followed by realisation of
their counterparts fabricated using transparent ITO as the emitter contact. The measured
electrical (d.c.) properties of the HBTs were then analysed. At first a comparison between
the opaque and transparent set of devices was made; this was followed by a comparison
between the HBTs fabricated from the different material systems. This electrical performance
analysis was followed by an extensive study of the suitability of these devices as phototransistors
[232]; a very brief demonstrational work was performed
to compare the results of vertical illumination versus edge illumination.
One of the consistent findings from the electrical study was that the emitter series resistance
for the ITO emitter contact devices was higher than those with conventional metal contacts.
This directly affected a number of other parameters of the transparent emitter contact
transistors such as the offset voltage which was higher compared to their conventional
counterparts. Further investigation revealed that the high series resistance could not be
attributed to the resistance of the ITO layer or to the cap layer resistance alone. It also
became apparent that the total contact resistance of the conventional devices also had other
resistive components, albeit much less significant in comparison to the r.f. sputtered contacts.
The additional resistance in the transparent contact must therefore be due to some sputter
induced damage to layers deeper than the cap layer. It was also observed that the thin layer of
In helps reduce this effect because preliminary devices without this intermediate metallic film
had even larger Ree'. More detailed study is necessary to investigate this source of resistance
and further reduction in the emitter resistance may be achieved by using a thicker layer of
indium without causing significant reduction to the transmittance of the ITO contact.
A study of the Gummel plot showed that the overall reduction in Ic vs. Vbe in comparison with
the metal contact devices again points to the likelihood of possible damage caused to the
emitter layer during r.f. sputtering which was not fully recovered during the ohmic contact
annealing. This deep damage results in an increased base ideality factor suggesting increased
SCR recombination at the base-emitter junction and also possibly greater base bulk
recombination; As a result, the measured d.c. gain, b is
consistently lower for the ITO emitter contact devices.
A study involving the C-V measurements on both these sets of devices suggested that there
was a noticeable change on the doping profiles of emitter layers in the HBTs due to r.f.
sputtering and that this does indeed give rise to a donor like effect in the affected regions.
Recall, this is consistent with the findings of the work on the ITO/n-GaAs Schottky photo
diodes. Furthermore, as there is no significant change in the collector doping profile, it was
assumed that the sputtering effects are confined before the collector layer. Since the C-V
profiling technique cannot yield the doping profile of the already highly doped p+ GaAs base
region, it is not possible to make definite conclusions based on these measurements about the
effects of r.f. sputtering on that region.
A final feature of this work is the relationship between the effect of r.f. sputter induced
damage on the base-emitter interface and the depth of this junction from the surface of the
wafer. For example in the work with the InGaP/GaAs HBTs, it was seen that the base ideality
factor for ITO contact devices fabricated on a wafer with a shallower base of 2725 (this
figure being the sum of the cap and the emitter layers) is 1.25 in comparison to 1.09 of their
opaque counterparts indicating presence of sputter induced damage giving rise to increased
SCR recombination at the heterojunction; in contrast, the difference in the nIb between ITO
and metal contact devices is insignificant for the wafer with the deeper base of 5000. These
results were also consistent with the work on the AlGaAs/GaAs HBTs on two wafers with
slightly different structures. The depth of the base/emitter hetero-interface for the InP/InGaAs
was 4000 and thus the damage sustained is less than in devices with shallower base/emitter
junctions. Recall, that the effect of sputter damage was not noticeable for the base interfaces
deeper than 5000.
Like their InGaP/GaAs counterparts, the InP/InGaAs material system also offers the
advantages associated with selective etching during device fabrication resulting in uniform
device characteristics across the wafer; this desirable property is reduced in case of the
AlGaAs/GaAs HBT fabrication where no suitable selective etchants are available. The
InP/InGaAs HBTs fabricated in this study, a low common emitter offset voltage (measured
when Ib ~ Ic ~ 0) of 61mV and collector/emitter breakdown voltage of 5.5V were observed.
However, it was evident from the output characteristics that InP/InGaAs HBTs have rather
poor saturation characteristics with high output conductance unlike the complete saturation
observed in typical GaAs-based devices. This is due to the high multiplication factor and the
low breakdown voltage associated with the low band gap of the InGaAs material.
In order to compare the turn on voltages, Vt, of the AlGaAs/GaAs HBTs, with typical
InGaP/GaAs and InP/InGaAs HBTs, the collector currents of three devices with same
geometries but different material systems were plotted as a function of base emitter voltage
with Vbc = 0. It is seen that the InP/InGaAs HBTs exhibit a lower turn on voltage (0.2eV)
compared to GaAs-based HBTs (0.8eV). This clearly demonstrates the advantage of InP-
based HBTs for low power circuit applications, such as mobile telecommunications
[199].
From a comparison of the respective gain plots of the devices on different material systems, it
was evident that while the AlGaAs/GaAs b varied with Ic, the
InGaP/GaAs showed little change with respect to Ic and the InP/InGaAs gain is almost
independent of the collector current. This was attributed to the influence of various
recombination processes in the base-emitter junctions of these devices and also the lower
surface recombination velocity associated with the InGaAs material.
Optical output results showed that there is no degradation of the active substrate layers or the
device characteristics as a result of r. f. sputtering used in the ITO deposition. It was also
shown that a HPT can be controlled both optically and electrically or by a combination of both
sources representing the potential to use a single device for the simultaneous detection and
amplification of an optical signal as well as its subsequent coupling with an electrical signal in
a single device. Much of the optical characterisation work was based only on the InP/InGaAs
devices.
A notable feature of the output characteristic of the InP/InGaAs HBTs with ITO emitter
contacts was the significant reduction in the collector offset voltage optical output of the
photo transistor to 15mV from 60mV in its electrical mode of operation. This reduction is
related to the absence of the finite base resistance in case of the optical mode where no active
base contact is required.
In order to understand the relationship between the optical gain and the d.c. current gain, the
responsivity of the base-collector p-n junction diode was measured and compared with the
responsivity of the device operating under the three terminal device mode. A plot of the ratio
of the responsivity of the InP/InGaAs HPT to the responsivity of the base-collector photo
diode vs. optical power at l = 780 nm showed that this is
approximately equal to b, in agreement with conventional
analysis of photo transistor operation. Similar measurements were made for a set of
AlGaAs/GaAs HPTs and found to confirm the relationship between the optical gain, the
electrical gain b and the photo diode responsivity at a
given wavelength.
The optical characterisations also involved the measurement and analysis of the spectral
response of these novel HBTs with ITO emitter contacts over the relevant wavelength ranges.
The AlGaAs/GaAs devices showed an operational range spanning over
l = 450 to 900 nm while for the InGaAs/InP devices this was
from 800nm to 1700 nm.
In both cases, the long wavelength cut-off is determined by the absorption edge of the narrow-
bandgap base and collector. For In0.53Ga0.47As
this corresponds to a wavelength of approximately 1650 nm. At shorter wavelengths (< 950 nm),
the photoresponse is limited by absorption in the InP emitter as well as the ITO transmission and
the monochromator grating efficiencies. The dip in the spectral response at around 1400nm
corresponds to atmospheric absorption of radiation from the monochromator and is not a
reflection of the device characteristic. This atmospheric absorption in the wavelength range
1340-1450nm is most likely due to moisture and CO2. The suitability
of the InP/InGaAs devices for operation at 1310 nm and 1550 nm wavelengths was clearly
demonstrated.
Finally, initial experiments comparing the benefits of edge coupling versus top coupling
showed that up to 40% of the incident beam is lost in the former mode. Device structures
need to be redesigned to suit the edge coupling mode. Further work is required to provide
more quantitative analysis.
A spectral response model was developed for the first time to understand, analyse and finally
to optimise the performance range of these HBTs [233].
Device parameters such as the doping concentrations, the ITO and the semiconductor layer
thickness as well as the material properties such as the absorption coefficients,
a, the refractive indices, n, and the generation efficiency were
used from the literature for the purpose of simulation. A 100% collection efficiency was
assumed for any electron-hole pairs photo generated inside the depletion region or within a
diffusion length in the neutral material. The spectral responses of the ITO/n-GaAs Schottky
diodes and the AlGaAs/GaAs, InP/GaAs phototransistors are simulated and compared to
measured data for the first time. In all cases, very good correlation between the model and
the measured results were obtained [204].
The results presented in this thesis provide valuable insight into the use of optically transparent
ITO for use in a wide range of optoelectronic devices as well as the respective optimisation.
A list of publications arising from this work is presented in Appendix A.
7.2 ITO-based Microfabrication
7.3 Optically Transparent Schottky Diodes
This requires further detailed study to understand both the cause and the nature of these
defects.
7.4 Transparent Gate HEMTs
7.5 Optically Transparent ITO Ohmic Contacts
7.6 Optically Transparent HPTs
7.7 Spectral Response Model
1998: Shabbir A. Bashar (in accordance with paragraph 8.2d, University of London
Regulations for the Degrees of M.Phil. and Ph.D., October 1997). The Copyright of
this thesis rests with the author, and no quotation from it or information derived
from it may be published without the prior written consent of the author.
Title
Dedication
Acknowledgements
Abstract
Symbols
Contents
This page is hosted by Dr. Shabbir A. Bashar,
e-mail: shabbir@betelco.com