View
22
Download
0
Category
Preview:
Citation preview
Retrospective Theses and Dissertations Iowa State University Capstones, Theses andDissertations
1994
Fabrication and characterization of hydrogenatedamorphous silicon films, CVD diamond films, andtheir devicesHao JiaIowa State University
Follow this and additional works at: https://lib.dr.iastate.edu/rtd
Part of the Condensed Matter Physics Commons
This Dissertation is brought to you for free and open access by the Iowa State University Capstones, Theses and Dissertations at Iowa State UniversityDigital Repository. It has been accepted for inclusion in Retrospective Theses and Dissertations by an authorized administrator of Iowa State UniversityDigital Repository. For more information, please contact digirep@iastate.edu.
Recommended CitationJia, Hao, "Fabrication and characterization of hydrogenated amorphous silicon films, CVD diamond films, and their devices " (1994).Retrospective Theses and Dissertations. 10485.https://lib.dr.iastate.edu/rtd/10485
INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI
thesis and dissertation copies are in typewriter face, while others may
be from any type of computer printer.
The quality of this reproduction is dependent upon the quality of the
copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignment can adversely affect reproduction.
In the unlikely event that the author did not send UMI a complete
manuscript and there are missing pages, these will be noted. Also, if
unauthorized copyright material had to be removed, a note will indicate the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by
sectioning the original, beginning at the upper left-hand comer and
continuing from left to right in equal sections with small overlaps. Each
original is also photographed in one exposure and is included in reduced form at the back of the book.
xerographically in this copy. Higher quality 6" x 9" black and white
photographic prints are available for any photographs or illustrations
appearing in this copy for an additional charge. Contact UMI directly to order.
Jauixi
i \ I * 0 A Beii & Howe!! Information Company
300 f^Qrth 266b Rosd. .Ar^n .Arbor. Mi 43105-1345 USA
313/761-4700 800/521-0600
Order Number 9503570
Fabrication and characterization of hydrogenated amorphous silicon films, CVD diamond films, and their devices
Jia. Hao. Ph.D.
Iowa State University, 1994
1\ fl I X_/ JL T JL JL.
300 N. Zeeb Rd. Ann Arbor, IvII 48106
Fabrication and characterization of hydrogenated amorphous
silicon films, CVD diamond films, and their devices
by
Hao Jia
A Dissertation Subinitted to the
Graduate Faculty in Partial Fulfillment of the
Requirements for the Degree of
DOCTOR OF PHILOSOPHY
Department: Physics and Astronomy Major: Condensed Matter Physics
Approved:
±r n,arge. or lYiajor worK
For the Major Department
For the Graduate College
Iowa State University rviucS , j-Owca
1994
Signature was redacted for privacy.
Signature was redacted for privacy.
Signature was redacted for privacy.
ii
TABLE OF CONTENTS
GENERAL INTRODUCTION 1
PART ONE. FABRICATION AND STUDIES OF a-Si:H 2
I. INTRODUCTION 3
A. Historical Background 3
B. Basic Properties of a-Si:H and related materials 4
C. Hydrogenated a-Si 8
TN c-*- — TT 3 —. 1/ uj. J. u. ciiiiu vui. v.;^ cii xu
E. Scope of This Work 16
II. SAMPLE PREPARATION 18
III. SAMPLE CHARACTERIZATION 24
A. Thickness Measurements 24
B. Tauc Gap Measurements 24
C. Infrared Absorption 31
D. Electron Spin Resonance Measurements 37
E. Photo and Dark Conductivity Measurements 3 9
?. Secondary Ion Mass Spectrometry 3 9
G. Small Angle X-ray Scattering 42
T\7 -p-rrcTTTrpc Avn nx coTTccT/^v: crn
A. Electronic Stability 50
B. Microvoids and H Content 59
C. Anomalous Hydrogen Diffusion 65
REFERENCES 7 6
iii
PART TWO. FABRICATION AND CHARACTERIZATION OF CVD DIAMOND FILMS AND DEVICES 82
I. INTRODUCTION 83
A. Historical Background 83
B. CVD of Diamond Films 87
C. Growth Mechanism of CVD Diamond 94
D. Scope of This Work 97
II. SAMPLE FABRICATION 99
A. Deposition System 99
B. Intrinsic and Doped Sample Preparation 99
C. Light Emitting Diode (LED) Structure 103
III. SAMPLE CHARACTERIZATION 105
A. Thickness Measurements 105
B. Scanning Electron Microscopy 105
C. X-ray Diffraction Measurements 108
D. Electron Spin Resonance (ESR) Measurements 110
E. Electroluminescence Measurements 113
T7 TV \Tr> C"O OV*" r» x V » K i l o u x j x o L / - L O V - . V J O O _ l - L O
A. Electroluminescence 118
B. The Native Defects of CVD Diamond Films 121
REFERENCES 13 3
GENERAL CONCLUSIONS 136
1
GENERAL INTRODUCTION
Semiconductor materials of hydrogenated amorphous silicon
(a-Si:H) and diamond films were fabricated by radio frequency
sputtering technique and chemical vapor deposition technique
respectively. As the nature of the two materials are not
closely related to each other, this thesis is arranged into
two parts.
The first part of this thesis described the studies of a-
Si:H films by infrared absorption, small angle x-ray
scattering, electron spin resonance, and secondary ion mass
spectrometry. The optoelectronic properties of a-Si:H films
were improved dramatically by annealing at 250-330°C; strong
correlation between hydrogen content and microvoid was found
in a-Si:H film; and contradictory evidences to the widely
invoked "multiple trapping" mode], for hydrogen diffusion in a-
Si:H were also found for the first time and discussed in this
work •
As the second part of this work, thin diamond films and
their related devices were fabricated and studied. The deep
ciodss V7s2rs siicccssfviilv
fabricated. A new defect center in the CVD diamond was also
identified for the first time through the hyperfine splitting
in the electron spin resonance spectra.
2
PART ONE.
FABRICATION AND STUDIES OF a-Si:H
3
I. INTRODUCTION
Unlike crystalline silicon (c-Si) which is well
understood and highly developed, amorphous silicon (a-Si) was
believed to be undopable and hence useless, until Spear and
LeComber^ successfully doped it both n- and p-type in 1975.
Since then, a-Si has been intensively studied and become
increasingly important in many advanced technologies such as
solar cells, thin film transistors, image scanners,
electrophotography, optical recording and gas sensors.
The early work on a-Si was carried out by Chittick,
Alexander, and Sterling^ in 1969. Their films were deposited
by glow discharge of silane (SiH4). However they failed to
produce either n- or p-type doped film due to the large number
of dangling bond states which effectively pinned the Fermi
>~ 7 4— "1 /J T T f •«-. J— . J — — ^ -A.V « a. nCiO i, 1W JL^/'« UliO. L- J_)CWX&
and his collaborators^ discovered the role of hydrogen in
hydrogenated amorphous Germanium (a-Ge:H). In 1976, Paul et
al^ also demonstrated that in rf sputtered a-Si:H hydrogen
passivates almost all of the unsaturated dangling bonds and
thus dramatically lowers the mid-gap density of states. The
Fermi energy can then be moved in either direction by doping.
4
B. Basic Properties of a-SirH and related materials
The periodic arrangement of atoms in crystalline
materials lead to Bloch's theorem and the relationship between
the allowed energies of electrons and their wave vectors.
Crystalline semiconductors are characterized by gaps of
forbidden energies separating the allowed energy bands. In
amorphous semiconductors, which are highly distorted systems,
the long range periodic order is erased. However, short range
order (SRO) still exist, as demonstrated by, e.g., Temkin and
coworkers^ and shown in Figure 1.1. The radial distribution
function (RDF) J(r) =47rr2p(r) is defined as the average number
of atoms in a thin spherical shell of radius r and thickness
Ar with the center of the shell on an arbitrary atom. p(r) is
the atomic density. By comparing the RDF with crystalline
germanium (c-Ge) in Figure 1.1, it is clear that the nearest
neighbor atom distance in amorphous germanium (a-Ge) is as
well defined as in c-Ge. However, the peaks become
increasingly sm.eared as the distance r increases since there
are no long range atomic correlation. The RDF also yields the
bond-bond angle which is given by
where r^ and X2 are the distances of nearest and next nearest
neighbor atoms respectively.
'•d p.
J { i ) (ilomj A*')
w a 'Tl !/)
o i-ii
0 1 o (D
(U ;3 ii.
\r
iu I o (D
l-t) ()
!•( (0 t-h
crj
tr o
hj
>•• — •w o
J(rl (ilorns A"')
UH O o
:>
O
00
O
6
One of the most important consequences of the lack of
long range order in amorphous semiconductors is that the wave
vector k becomes meaningless and hence the Bloch theorem is no
longer valid. However, energy bands and band gaps still exist,
and a density of states N(E) can still be defined to describe
them. Due to the distortions of the network, the sharp
features in the density of states are smeared out (Figure 1.2)
by localized tail states generated from the bond strengths
variation. These tail states are critical in controlling
electronic transport, as almost all earners rapidly
thermalize to tail states at the band edges. The deep
localized states in the gap arise from coordination defects,
predominantly Si dangling bonds, which determine many
optoelectronic properties by trapping and recombination.
A dangling bond arises, for example, when a silicon atom
bonds to only three other silicon atoms. This may happen at
surfaces where another Si atom required for bonding is simply
missing. Or, it may occur in the bulk as a result of the
distorted Si network. In either case, this leaves a Si sp^
orbital unpaired to another sp^ orbital. Thus, a neutral
dangling bond results in an unoccupied near mid gap state
(Figure 1.2).
7
Conduction bond extended states
iij
>* CP
c LiJ
T3 dangling bonds
Valence band
Localized conduction band tail states
T3 dangling bonds
Localized valence band ^^tnil states
O \^AlOltUO^ OIVJIOO
Density of States, N(E)
Fig. 1.2 Davis-Mott model of the electronic density of
cKrM.T-ir^rr W.- Vm M A * A A O N-I' A k f A • s»/ WAA V2.
band, conduction band, tail states, and dangling
bond states near i?.id gap
8
C. Hydrogenated a-Si
Due to the large amount of dangling bonds (up to
lO^O/cm^) , the gap states largely mask the gap in a-Si thus
pinning the Fermi energy level at the center, making doping
impossible. However, the incorporation of hydrogen in a-Si, to
yield hydrogenated a-Si (a-Si:H) greatly reduces the density
of dangling bonds and sharpens the band tail edges (Figure
1.3), enable doping and dramatically improving the
optoelectronic properties.
The reduction of the density of dangling bonds is a
result of hydrogen passivating the dangling bonds, and can be
seen from the electron spin resonance (ESR) measurements.
Typically, spin density is about lO^^-lO^O cm"^ in a-Si and as
low as 10^^ cm~2 in device quality a-Si:H.
Addition of hydrogen also increases the energy gap of a-
Si and produces sharper band edges. The energy gap or Tauc
gap (Eg), which can be measured from optical absorption,
increases continuously from 1.4 ev to 2.0 ev with increasing H
content and is attributed to the silicon-hydrogen alloying
effect. The tail states are mainly formed fromi the weak Si-Si
bonds attributed to the variations of the local bond lengths
and angles. The addition of hydrogen breaks these weak Si-Si
bonds to form stronger relaxed Si-H bonds, thereby reducing
the width of the band tails.
9
22^
Amorphous Si containing no hydrogen
U
O O
containing hvdros,en
Valencc I j Conduction
band Forbidden band
Fig. 1.3 Effect of hydrogen addition in amorphous silicon
10
D. Electronic Stability and Hydrogen Motion
(i) The Staebler-Wronski Effect
One of the major applications of a-Si:K is for solar
cells, which are inexpensive in comparison with c-Si cells and
can be fabricated on a large scale. The problem limiting the
application of a-Si:H solar cells is the so called Staebler-
Wronski effect (SWE). In 1977, Staebler and Wronski^/^
discovered a reversible light induced degradation process in
THo p^otcconciuctivity cc^ciu.ctl^ V^ "ty C'q
of glow discharge (GD) deposited a-Si:H were reduced
considerably upon illumination with intense light. While
was reduced by factor of eight, decreased by four orders of
magnitude (Figure 1.4). Upon annealing at 150°C for four
hours, both ap^ and recovered their original values and the
entire process could be repeated.
Xt IS generally agreed that the illumination itself does
not create deflects in tne micigap regionj ra*ch3r some eiec"cron~
V> I VN T O O T 1 -N 4- <-s "v— -> * 2. AO
jTSCoiulDins.ion/ s.n,v5. "tlis irsj.Ga.SGc3. Gnsir^y cir£a.u.GS tliG ciGfGC't.s
which degrade the optical and electronic properties. Stutzir.ann
et al.^*^ carried out a detailed study on the SWE and reached
the following conclusions: (a) The SVJE effect is basically an
intrinsic bulk effect, independent of the concentration of
major impurities like oxygen and nitrogen, (b) The generated
11
in-3
10 -4
10 -5
I E o
10 -5
^ Illumination
> 10 — I s I
O 10 -s
1 A-s[_
e \
r — 1 0 •lU • L
I
ir_ 0 50 100 150 200 250
Time (min)
Fig. 1.4 Staebler-Wronski effect
12
defects have an ESR signature identical to that of dangling
bonds, (c) The generation rate of the metastable defects does
not depend on the incident photon energy in the range between
1.2 and 2.1 eV. This suggests that the defects are created
after the thermalization of the optically excited carriers
into the deep band tails. However, considerable controversy
still remains as to the nature of these defects and their
generation mechanism.
(i.X) Hvur'ocrcii mouxon
The incorporation of hydrogen in a-Si leads to many
beneficial effects as mentioned earlier. Therefore, hydrogen
motion in a-Si:H could be crucial in determining much of the
electronic properties of the material. The weak Si-Si bonds
broken by the released energy from the same nonradiative
recombination events as in the SWE might require a hop of an H
atom from a neighboring site to stabilize this newly formed
defect and this may constitute the basic diffusion step for
hydrogen diffusion.Kence the SWE may be viewed as "light
enhanced hydrogen diffusion". This potential role played by
hydrogen in the origin of the metastability has provided an
impetus to study the hydrogen diffusion in these materials.
The ability of hydrogen to move into, out of, and within
a-Si:H has both beneficial and undesired results. The low
defect density is a beneficial result from H termination of
13
dangling bonds. Hydrogen is, however, also apparently
responsible for the instability of a-Si:H at elevated
temperature due to the increasing rate of creation of dangling
bond by hydrogen evolution.
We initially assume that H motion obeys Pick's second law
of diffusion
dC = D—Tr 1,2
dt dx 2
where C(x,t) is the hydrogen concentration and D is the
diffusion constant.
The closest form of boundary conditions for the
multilayer structure is that of semi-infinite slabs in contact
at the interfaces (Figure 1.5). The boundary conditions for
such a structure are: C(x,0)=Co for x<0, C(x,0)=0 for x>0, and
/I / 0- \ A nTU 4 i_T a_ - v.. -1 i- J — V/. X no OWJLC1U.JLW11 UllC «^XJLJL«-1AXWA1 CTMU-dUXWll UW
these boundary conditions has the form of a complementary
error function^^
C(x,t) = C o / 2 erfc[x/(4Dt)1 / 2 ] 1 . 3
where CQ IS the deuterium concentration at boundary (x=G); t
is the annealing time. The complementary error function is
defined as
14
C
I X
Fig. 1.5 At t=0, C=Cq to the left of x=0. After diffusion
has proceeded for a period of time (as shown by
solid line), a finite amount of D has diffused into
15
c» _ 2 erfc{x) = \ e ^ dy 1-4
;c
For disperive diffusion where D is time dependent, Dt is
replaced by
r
0(i) = jD(T)dT 1.5
therefor, the solution can be expressed as
C(x,t) = Co/2 erfc[x/(40)1/2] 1.6
despite the time dependence of the diffusion constant, the
profile generally is still an error function, as verified by
SIMS.
The time dependent diffusion constant of hydrogen in a-
Si:H was found to obey a power-law given by^-"!^
" V - /
where Dqo is a prefactor, co is the attempt frequency and a is
the temperature dependent dispersion parameter which is
generally greater than zero since diffusion normally slows
down as time increases. The "hydrogen glass" model, proposed
16
by the Xerox group,suggested that a depends on the
temperature T as
a = l - P = l - T / T q 1 . 8
where IcTq is the width (decay constant) of the exponential
distribution of multiple trapping (MT) sites.
E. Scope of This Work
In this work we studied the electronic and optical
properties of a-Si:H. The samples were prepared by rf
sputtering at varying rf power and H content. Thus the
microstructure of the samples could be controlled and the
relation between hydrogen content, microstructure, electronic
and optical properties could be studied. The major techniques
used in this study were infrared (IR) absorption, secondary
^ ^ ^ ^ -.IT ^ ^ I-*-.---i.v AA WA. jr f j\. l\Q. y O wcx u. a
(SAXS); and electron spin resonance (ESR). The IR yielded
information on the total hydrogen content and the nature of
hydrogen bonding; the SAXS provided insight on the microvoid
structure and content; the ESR yielded the dangling bond
density, and deuterium SIMS profiles of a-Si:H/a-Si: (H,D)
^ i ^ - v — / - > • > - » V > ^ f r O v - / — o V a. ^ =.LA wjL. AiiQ Wii L4.X. Ci i jL J. X wi i •
a-Si:H and a-Si:H/(H,D) multilayers of varying thickness
were prepared at different rf powers ranging from SOW to 600W.
17
The hydrogen partial pressure and the sputtering power
determined the total hydrogen content which varied from 2 to
3 0 at. %. The target to substrate distance for most of the
samples was kept at 1.25" or 2". The substrate temperature was
controlled from room temperature to 250°C. The annealing
temperatures were 150-430°C, and all annealing processes were
carried out in evacuated pyrex tubes.
18
II. SAMPLE PREPARATION
As mentioned above, all of the samples used in this work
were prepared by reactive rf sputtering at a frequency of
13.56 MHz. A schematic diagram of the deposition system is
shown in Figure 2.1. Sputtering is one of the most common
techniques for fabrication of various thin films, rf sputter
deposition of a-Si:H is accomplished by bombarding a Si target
with energetic (up to several kev) ions of an inert gas (Argon
in this work). The inert gas is ionized and accelerated
towards the target by the self-biased electric field. Those
ions then strike the target surface and knock out atoms from
the target. These atoms then travel across the plasma and
deposit on the substrate.
The target used in this work was a 6" diameter 9 9.99%
pure polycrystalline silicon target, mounted on a water cooled
stainless steel backplate located at the top of the plasma.
Several types of substrates were used in this work for various
purposes. Double side polished single crystalline silicon
wafers were used for IR absorption and SIMS m.easurements,
Corning 7095 glass slides for UV-Vis-NIR absorption and ESR
measurements, and thin aluminum foils for SAXS and nuclear
magnetic resonance (NMR) measurements. The substrates were
placed on a heated plate directly underneath the target and
the distance between the target and the substrate holder
19
Generaiion
Matching
Target
Ti-Ba!l Subiimlnator
Ggs Mixing Tank Sputter Etch
Butterfly Valve Pedestal
Argon
- n Hydrogen i-CiO-" •
HDopant Gas
^2)— Nitrogen
npxp Gate Valve
Turbomolecuiar Pnmn " r"
Fore Pump
Fig. 2.1 Schematic diagram of the rf sputtering system used
for deposition of a-Si:H
20
(heater) could be easily adjusted from 1" to 2". The
background vacuum pressure prior to film deposition was pumped
down to about (3±l)xl0~'^ torr. During deposition the gate
valve connecting the plasma chamber to the turbo pump was
opened three turns to slow down the pump rate. Gases were
introduced at a preset partial pressure; the flow rates were
controlled by micrometer valves. The rotatable pedestal in the
chamber enabled cleaning of the target by presputtering.
After the desired gas flow rates and partial pressures
are achieved,, rf power is applied to the target while the
pedestal, on which the substrates were held, is grounded. The
strong rf field then ionizes the gas atoms and forms a plasma.
When the target is at a negative potential the positive ions
in the plasma are attracted to the target. Because the
potential of the target alternates at a very high frequency
(13.56 MHz), the positive ions are not able to reduce the
negative porenriai of the target to zero before the rf power
2.2. to 2. positivs VJlisr* is
s. positive pot^3.sttZTwCts tlis siscwircns in
plasma. Since electrons are much lighter and hence faster than
ions, they are much more effective in reducing the positive
bias of the target, and the average (DC) potential of the
target is reduced from zero to some negative value. Figure 2.2
shows the rf voltage supplied by the power source, V-, and the
potential at the target, Vj^. This naturally occurring self-
21
Time (arb. units)
Fig. 2.2 Voltage vs time characterization for source voltage
V^, and the target voltage V-^, in a rf sputtering
system (From ref. 19)
22
bias of the target causes the positive ions in the plasma to
bombard the target surface and knock out atoms of the target
which are then deposited on the substrate.
Keating the substrate during the deposition process
generally has a beneficial effect. It can enhances the lateral
mobility of the surface atoms thereby reducing an incipient
columnar morphology and microvoid content of the films. At
-250Oc the di- and tri-hydride configuration on the growing
surface tend to break and form molecular H2 which is removed
from the film. Therefore, heating the substrate will reduce
the density of those bonds which normally create voids or
surface-like defects. However, when heating above -400°C will
break the mono-hydrogen bonds and increase the density of
dangling bonds.
Multilayer films of a-Si:H were also prepared by
sandwiching an a-Si:(H,D) layer between two a-Si:H layers.
Figure 2.3 shows the structure of a multilayer film, and an
ideal as-deposited SIMS profile for such a structure. By
sandwiching an a-Si:(H,D) layer in a-Si:H, we can monitor
hydrogen-deuterium interdiffusion through SIMS measurements.
As the major difference between the diffusion of deuterium and
hydrogen is presumably due to their different masses, their
diffusion mechanisms are expected to be identical.
23
a-Si:(H,D)
a-Si:H j — ^1
Si substrato
o q) in \ yj 4-1 C 3 o r i
cc .C cn
•r-i yj
cn s r-t cn
>SI:H a-Sl:{H.D) a-Sl:H
Depth
Fig. 2.3 The multylayer structure and an ideal SIMS profile
of a-Si:H film
24
III. SAMPLE CHARACTERIZATION
Various techniques were used to characterize the
structure, optical, and electronic properties of a-Si:K films.
We used a Dektak profilometer to determine the thickness of
the films, optical absorption to determine the optical band
gap (or Tauc gap), IR absorption to determine the bonding
structure, SIMS for diffusion measurements, ESR for dangling
bond spin count, and SAXS for microvoid characterization.
A. Thickness Measurements
A Sloan Dektak profilometer was used to measure the
sample thickness. A stylus mechanically determines the
thickness of the films with an accuracy of ± 100 nm. During
the deposition, one piece of Corning 7059 glass substrate was
partially masked to create a step for the thickness
measurement. Figure 3.1 gives a schematic illustration of this
technique.
B. Tauc Gap Measurements
The optical band gap, or Tauc gap Eq, was determined by
using Cary 14A spectrophotometer. A schematic of the dual beam
instrument is shown in Figure 3.2. It can measure the optical
density (OD) from 300 to 2000 nm. The incident beam of
intensity IQ passes through the reference compartment while
fvietai Mqsk Z
- J. J L / /
t TJL F
Glass Substrate
Rofn i \j . . ro qni litprlriri i^iv^ v-yi^umciuivj i ^
1
i
y
Dektak Stylus
= -}Film Thickness
i >
U WZZZZZlr
•a-Si :H
Gloss Substrate
After Sputtering
Schematic representation of thickness measuremen
26
the transmitted light of intensity I exits through the sample
compartment. The width of the entrance slits of these
compartments is automatically adjusted to balance the
intensities of the two exiting beams. The optical density is
defined as
OD = logio(Io/I) 3.1
When light passes through a uniform homogeneous film of
thickness d, its intensity will be attenuated
I = Iq exp(-ad) 3.2
Since the film is deposited on Corning 7059 glass for
mechanical support, the reflection at the interfaces of
air/film and film/glass must be taken into account. The
-r> O rp T I T VN ^ /-> V^v-N-V- T" T" O
as 19
r \
- ad) cos{A mi.-d I ?.) ^ ^
where n a=l, n f , and n g = 1.53 are the indices of refraction of
air, film, and Corning 7059 glass, respectively. For a-Si:H
film, the accepted value of nf is -3.50, but it varies with H
content and wavelength. In Ecj. (3.3) , TQ is the transmission
27
Tung«t»n Lomo
u I nmn 1^9 1 • i! 1
9--
Photo Tul>«
A Ht Lamp
B,C>i Lef»«s D Bntronce Slit
r mkrofs H intwrmdJot# Slit
L Exit Slit
P 60 cpft Rototing
^tnmr
Fig. 3.2 Optical system of Gary 14A spectrophotometer
28
coefficient of Corning 7059 glass, which has a value of 0.97
in the wavelength range of interest.
Figure 3.3 shows a typical absorption spectrum. The
interference fringes in the wavelength region where a is small
are due to the cos(Aunfd/?L) term in the denominator of Eq.
(3.3). However, when a become larger at short wavelength, both
terms in the denominator of Eq. (3.3) become so small that we
can neglect them, and a is then given by
1 , . 10"°° a = —mi 1 3. a
The optical gap Eg can then be determined by using the
relation suggested by Tauc et. al.^C
{ccnjhco)^'' = B{jico — E^) 3 . £
where B is a constant which depends on the density of band
tail states. A Tauc plot is shown in Figure 3.4. The optical
gap Eg is obtained by extrapolating the region which obeys Eq.
(3.5) to a = 0; the intercept of the line with the fico axis
yields Eg.
In amorphous materials the optical properties are
characterized by three different absorption regions: At high
incident photon energies, the interband transitions are
similar to those which occur in single crystals. Below this
29
/ uu • , o ( 0\J
r~i r\
ouu /—\ R- Y—V gou yuu
Wavelength (nm)
3 A typical Gary 14 output, optical density vs.
wavelength
30
200 n—i—i—i—r
y'
a o
150 y -v
y
0)
cv
*
100
1 .• i .d I . ( i.O
Photon Energy (eV)
Fig. 3.4 Typical Tauc plot for optical gap
31
region, absorption is due to transitions between the band tail
states which are absent from crystalline materials; these were
first observed by Urbach^l in ionic crystals and are known as
"Urbach edge transitions". Below the Urbach edge, absorption
is due to transitions involving the defect states in the gap.
C. Infrared Absorption
Infrared (IR) absorption is an important technique that
is widely used in characterizing a-Si:H. It can provide
information on the total H content, its bonding
configurations, and oxygen or nitrogen contamination in the
films. The IR measurements were made on samples deposited on
double-side polished single crystal Si wafers by using a
single beam IBM model IR98 Fourier Transform infrared (FTIR)
spectrometer. The absorbance spectrum of a blank substrate was
subtracted from the absorbance spectrum of the substrate-film
sample to give the absorbance spectrum of the film alone.
A typical IR spectrum of a-Si:H is shown in Figure 3.5.
Three main Si-H vibration modes were first identified by
Brodsky and coworkers.22 These are (1) the Si-K stretching
mode at 2000 - 2100 cm~^, (2) the Si-H2 and Si-H3 scissors or
bond bending mode at 840 - 890 cm.~^, and (3) the Si-H wagging
mode at -640 cm~^. The corresponding bonding configurations
are depicted in Figure 3.6. The Si-bonded H content and Si-H2
32
configuration density can be calculated from the 64 0 cm~^
wagging mode by using the relation:
Nff = Aj a{cL>) Id) da ^ ^
where A is a proportionality constant which depends on the
oscillator strength and a(o) is the absorption coefficient at
frequency co. The values of A for different vibration modes
were determined by Shanks et al.23 Cardona ' / 25
given in Table 3.1.
The 640 cm~l band, which shows no discernible structure,
is attributed to the wagging mode of Si-H bonds in any bonding
configuration (see Fig. 3.6). The prefactor A of this mode has
been found to be nearly independent of both deposition
conditions and total H content calculated from this
band is thus considered to be reliable. From Eq. (3.6) and the
experimentally determined A value (Table 3.1), the Si-bonded H
content can be expressed as
H.X.(at, %) = 1.125 As4o/d 3.7
where Ac/n is the area (in cm~^) under the absorption peak of
640 cm~^ wagging mode and d is the film thickness in microns.
Similarly, the concentration of bonds corresponding to the 840
33
2SEB vsyes'jxsess ck-:
Fig. 3.5 Typical IR absorption spectruxn of a-Si;H
34
Table 3.1 The various vibration modes of hydrogen in a-Si:H,
their bonding configurations, and the corresponding
proportionality constant A (from Ref. 23; see Eq.
(3.7))
Uavenumber 540 8A0-890 2000 -2100 2100 ( cn
Mode Wag Sc i sso rs S t re t ch S t re t ch S t re t ch
Bond ing S i -H S i -H2 S i -H S i -H S i -H2
S i -H2 S i -H2 ( I so la ted ) (C lus te r ) SI -Kt Si -H3 (m ic rovo ids ) ^ (m ic rovo ids ) ^
A ( cm-2 ) 1 .6x10^9 2x10^0 2 .2x101^ l .Tx lO^O 9 .1x l0 l9
^Compressed m ic rovo ids
' ^H i c rovo ids v i t h rad ius ^ 2A
See t ex t f o r de ta i l s .
35
•Si OH
t Vxvx?
i 9 e « a ' SYM SYM ASYM
• ASYM, u cu cj 3
1 2 -3
2000cm"' 2090cm*' 2i20cm'i
9r> q''9> a
A f T SYM ASYM SYM
I r J
(11° 3
390cm-' SSOcm-i
'r>
v- r i / \ / i \ n
• ® » cj}^ , yi w
O < I I I J
w cj"
64 0cm"' 590cm''
Fig. 3.6 Illustration of vibration modes of stretching
(top), bending (middle), wagging and rocking
(bottom)
35
cm~l Si-H2 and Si-H3 scissors mode vibration can be expressed
as
n(j(at. %) — 10.44 agaq/'*^ 3.s
The stretch mode at 2000 cm~^ is attributed to isolated
Si-H bonds in the bulk. The peak, at 2100 cm~^ has been
attributed to silicon associated with di- and tri-H bonds or
to mono Si-H at internal surface of larger microvoids.^2
However,, it is not possible to distinguish between silicon
bonded to two or three hydrogen atoms from either the scissors
or stretch peak. It has been pointed out that the stretch
frequency of mono Si-H bond shifts from 2000 to 2100 cm~^ as
the radius of the microvoid around the hydrogen increases to -
2 A.25-27
Presence of oxygen at levels above -0.5 at. % in the
films can also oe easily aetected from rne IR specrra. Tne 900
c~;—— p3=k is related to bulk Si~0 bonds, while ths 1100 cm~~
psH}c is stinri-biitsd. to tlis Si~0 tond. intsirnBl su^rfscss- Tins
later is often observed in films exhibiting columnar
morphology, in which case it grows with time upon exposure to
air.
37
D. Electron Spin Resonance Measurements
Electron spin resonance (ESR) measurements were used to
measure the density of unpaired dangling bonds in a-Si:H
films. The measurements were performed on a Bruker ER2 2 0 DSR
X-band ESR spectrometer. The samples were deposited on Corning
7059 glass cut into pieces of half inch by three millimeters.
The spin density and the g factor were obtained by comparing
the ESR spectrum with a reference DPPH of known spin density
and g factor.
An electron will interact with a magnetic field due to
its magnetic dipole moment The interaction will cause an
energy level splitting or Zeeman shift
the dipole moment is related to the electron spin S by
where g is the g-factor and P = eTi/2mQ is the Bohr magneton.
Since the eigenvalues of S along the magnetic field direction
E = -}l-B 3 . 9
\l = -gps//z 3 .10
^>*0 / O +*"H^ *7 O CiTTi Y-V C"V>1 — — • - / — r — - - — w — ' ^ f r
ae = gpb 3 .11
38
Resonant transitions between these levels can now occur if an
electromagnetic field of frequency co given by
no)= gpB 3.12
is applied normal to B. For B = 3340 Gauss and g = 2, the
resonant frequency is o = 9.3 GHz (X-band).
The ESR measurements were performed by inserting the
sample in a microwave cavity placed in a magnetic field
(Figure 3.7). The microwave frequency was set at 3.3 GKz and
the magnetic field was swept from 3330 to 3350 Gauss. A small
modulation field of 4 Gauss was applied parallel to B and
referenced a phase sensitive lockin detector which then
yielded the resonance signal detected by a microwave bridge
attached to the cavity. This method yields the first
derivative of the absorption spectrum, as shown in Figure 3.8.
The total number of spins in the sample can be calculated
by comparing with a reference's
where the superscripts r and s denote the reference and sample
respectively. In Eq. (3.13), Ah^Q is the derivative peak-to-
peak width of the resonance spectrum, Y is its amplitude,
is the amplitude of the modulation field, is the microwave
39
power, A is the line shape factor, G is the lockin gain, and
Np is the number of scans.^8
E. Photo and Dark Conductivity Measurements
Conductivity measurements were performed by using a home
made two-probe contact setup which is similar to the four-
probe method^^ and had been calibrated with a known sample.
The photoconductivity was measured with a heat-filtered
tungsten-halogen lamp which provided an intensity of 200
mW/cm^ on the sample surface. The photo- to dark- conductivity
ratio was then calculated and compared among samples to
characterize the optical performance.
F. Secondary Ion Mass Spectrometry
Information on long range hydrogen (or deuterium) motion
can be monitored by using secondary ion mass spectrometry
(SIMS).^'^ This technique is often used to profile the near
surface region of a solid as a function of depth. It uses a
primary ion beam to bombard the sample surface and sputter off
atoms and ions from the sample. Secondary ions can then be
identified by a mass spectrometer.
A schematic of the Perkin Elmer SIMS model PHI 5300 SIMS
used in this work is shown in Figure 3.9. A primary beam of
positive ions (Cs''") with diameter -50 S-im and energies -5 keV
wss s 2.^ 30^ ^ s
40
Magnet
Microwave
Generator
Sample
Fig 3.7 Schematic of ESR system
41
D <
b>
'A ul
3.0
2.0 -
1.0 -
0.0
-1.0
-zd
3.29 3.31 3.33 3.35 ilhousanda!
Magnetic Field iKGau&s)
Fig 3.8 Typical ESR spectrum of a-Si:H
42
beam was used to raster a 500x500 jim^ area on the sample
surface. The secondary ions from the sample surface were then
analyzed by the SIMS analyzer which consisted of an energy
filter, a quadrupole mass filter, and a signal amplifier.
Figure 3.10 shows a typical deuterium depth profile of an
a-Si:H/a-Si:(H,D) multilayer film. The x axis corresponds to
the sputtering time, which is proportional to the depth. The y
axis is the secondary ion current intensity in units of
counts/second. The sharp drop in the concentration is used to
define the interface. It also determines the depth resolution
of the profiles, which is typically - 150 a at the interface
of a - 1 um thick film.
Besides hydrogen and deuterium, other atomic species such
as oxygen, carbon, silicon, etc., can also be monitored. This
is achieved by simultaneously monitoring their corresponding
m/e values. Since the quadrupole mass analyzer does not employ
magnets, this peak switching process can be done rapidly
without hysteresis effects.
G= Small Angle X-ray Scattering
Microvoids in a-Si:K and its alloys have been the subject
of extensive research in recent years. A number of methods^
have been used to infer the existence of microvoids and
attempt to quantify them, but small-angle X-ray scattering
(SAXS) remains the most direct, as it can provide information
43
ION GUN
\
\ SPUTTiiR
"^^angli: SIMS
ANALYZER PRIMiARY
IONS SECONDARY
IONS ELECTRON
MULTIPLIER
VACUUM
T C T) T • ".*
Fig 3.9 Schematic of SIMS system
44
as deuos i t ed
i i i 1 i i i i i i i i i i i i i 1 i 1 i 0 .0 ' 2 ,8 5 .6 8 ,4 1 1 . 2 14 ,8 16 .8 19 -6 ' ^2 .4 25 .2 28 .0
SPUTTER TiliE, lilH.
i i 1 i 1 i i i i i i i i i i i i i i i
0 Ui to
M
t
t-" d 0 u
y
0 0 J
723 hrs @300OC
"A*--V
i i . i i I I ! I
e .9 i f i 6 .n 8 14 ,s i ? . s
SAUTTER TiliE, iliH. '•/a 0 0 i.'J t -J V I V
i I I 0 .ri n
1. 'v* I V i ; 0
Fig 3.10 Typical SIMS depth profile of a-Si:H/a-Si:(H,D)
45
on the microvoid size, shape and number density. Small-angle
electron scattering (SAES)'^° and SAXS'^^"'^'^ were applied
several years ago in studies of unhydrogenated a-Si and a-Ge,
where observations of density deficits were explained by the
presence of internal -5 nm sized voids. Subsequent studies of
a-Si:H using SAXS'^^"'^'^, SAES^^"^*^, and small-angle neutron
scattering (SANS)^^"^^ all demonstrated the existence of
microvoids in non-device-quality material.
The a-Si:H films studied in this work were deposited on a
10 thick iron-free aluminum foil, which was then cut into
eight strips after deposition and stacked to produce
sufficient thickness for the SAXS measurements. These
measurements were carried out with a Kratky Compact Small-
Angle System (Anton Paar, Graz, Austria) attached to a Rigaku
rotating anode (Cu) X-ray generator. Some measurements were
made with the sample normal tilted at an angle to the beam.
To estimiate the microvoid volume fraction, Vf, the
foliov;ing integral expression, which is valid for a simple
two-phase system and for the line collimation geometry, is
used^^
\hi{h)dh = 3.14
where 1(h) is the SAXS intensity. h=27c(20)/>. is the magnitude
of the scattering vector, and Ao is the difference in the
46
electron densities of the two phases. The quantity K depends
on various geometrical factors, the electron scattering cross-
section, the X-ray wavelength, and the absorption coefficient
of the filn. Because of the uncertainties in the geometrical
factors, the theoretical value of K was ignored in favor of
the determined quantity. To apply Eq. (3.14), first an
experimental value of K is obtained for an a-Si:H sample by
assuming that the SAXS and the density deficit determined by
the flotation method are caused entirely by microvoids. This K
is then used to predict Vf for all other samples.
Figure 3.11 displays the SAXS data (I*h vs. h) of samples
#1-5 as deposited. The radius of gyration, Rg, which
characterizes the size of the electron density fluctuations,
can be obtained from the linear behavior of Guinier plots^®
lQi) = I txv{-Rlh- n) 3.15
The departure from linearity indicates that the films contain
more than one size of microvoids.The strong scattering of
sample #1 indicates that irs Vf is larger than others. Sample
#2 had much stronger scattering that the intensity has been
scaled down by factor of 3 in Figure 3.11. Figure 3.12 shows
the tilting effect of sample #1 at 45° and 60°. The tilt-
dependence changes suggests that the microvoids in this film
are oriented, rod-like voids^'^. From previous SAXS studies on
47
it appears that its microstructure is columnar. The
other three samples (#3-#5), which contain -10 at. % Si-bonded
H and -0.5 vol. % microvoids, showed no anisotropy upon
tilting, suggesting either spherical microvoids or a random
orientation of non-spherical microvoids.
43
Fig 3.11 Plots of I*h vs h for the as-deposited a-Si:H
films. Note the intensity of sample #2 has been
scaled down by factor of 3
49
1.50
non-tilt in s
45° - t i l ted
0.<30 0
h (nm
•ig 3.12 0° and 45° tilted SAXS of sample #1, showing the
strong tilt angle dependence of Vf (see Eq. (3.15))
50
IV. RESULTS AND DISCUSSION
A. Electronic Stability
Fifteen samples were deposited on nominally unheated
substrates for the systematic study of electronic and optical
stability- The rf power Pj-f and H2 partial pressure Ph2 varied
from 200 to 600 W and 0.2 to 1.0 mT, respectively. The samples
were then characterized by optical absorption to determine
their optical gap Eg, IR absorption to determine the Si-bonded
K content and. its confiyuration, and. the photo— & dark.—
conductivity to characterize the optoelectronic performance.
Table 4.1 summarizes the properties of the samples.
generally increased with decreased with increasing Pj-f
(Figure 4-1). The optical gap Eg increased with the hydrogen
content. The photo- to dark-conductivity ratio for all
as-deposited samples was very poor, and the ESR measurements
confirmed the high initial defect density of -10^^/cm^- These
defects are recombination centers which decrease the
photoconductivity and pin the dark Fermi level near mid gap.
However, after annealing at 3 00°C for 9 hrs,
dramatically increased to -10^ (Figure 4.2). Further annealing
at 3 00°C showed slight additional improvement. But the crp>,/ajj
of some samples began to decrease following annealing at
51
Table 4.1 Characterization of the as-deposited samples
#
PH2
(m-c)
Power
(W)
Thick
(M-)
%
(ev)
CH
(%) (ncm)-l r
I r
1 200 1.15 1.78 22.2 2.1x10-10 146
2 1 300 0.81 1.69 21.7 3.0x10-10 • 5.8
3 1 400 0.85 1.67 20.3 9.4x10-10 5.5 j
4 1 500 1.00 1.55 11.7 4.0x10-9 3.2
5 1 600 0.95 1.64 10.7 2.1x10-8 1.5
6 0.5 200 1.10 1.62 14.2 6.5x10-9 1.3
7 0.5 300 1.10 1.58 11.2 1.2x10-8 1.2
8 0.5 400 1.10 1.55 7.1 !. 5.4x10-8 1 1 j t 1
j 9 j 0.5 j 500 j 0.94 1.47 V.5 j 9.3x10-8 i 1.1
10 j 0.5 j 600 1 1.00 1.53 6.56 3.6x10-7 L2
11 1 0.2 200 1.25 1.41 1 8.0 2.0x10-6 1.1 i 1
12 I 0.2 i
300 1.15 1 1.39 •i i
6.67 5.0x10-6 j 1-0 ! 1 > * ^ j ^ i i 1 ,
13 I 0.2 i 400 I 1.05 I 1.37 j 8.8 j 2.9x10-5 i 1.0 |
14 I 0.2 j 500 I 0.77 j 1.34 j 4.8 j 7.8xlO-S j 1.0 j
15 i 0.2 I 600 i 0.60 I 1.36 I 6.4 ! i ixlO-4 i 1.0 ! • • i ! ! } X. .L A. J. V,' , 1
52
20 n I i i i ' I i i i ' i I i I i I i > i i i I '
15 a o c o
10 d o w! o u
^ 5
ph2 = 0-5 mt
i 1 i 1 i i i \ \ i i t i i i i i i i i i i i
200 300 400 500 600
RF Power (W)
V-i o
25
20
1 I I I I I I > I j t 1 i I I 1 I 1 i I I i 1 i j I 1 1 ]
d 15 o o
^ id o -to >>
200 W
600 W
0 h i ^ i i i i i i i i i i i i i i i i i i i i i i i i i i i i ii t
0 • 0.2 0.4 0.6 0.8 1 1.2.
H2 Partial Pressure (mr)
Fig. 4.1 Correlation between deposition parameters and the
resulted hydrogen contents
53
200 W 200 w 200 W 300 W 300 \V 300 W
i mi \J.D mi vj.Z mi 1 mT 0.5 mX O.z mi
Fig. 4.2 Optical performance characterized by OQh/'Jd
dramaticallv chanced after annealincr
54
3 3 0OC for 10 hours.
Figure 4.2 also showed that higher Cjj and lower Pj-f
resulted in better CTph/c^d' which is understandable because
more hydrogen is available in the sample, and more dangling
bonds will be terminated by hydrogen during its motion,
reducing the defect states in the gap; higher P^-f normally
creates more defects due to the higher energy bombardment
during sample preparation.
The dramatic improvement of during the annealing
processes is believed to be related to hydrogen motion. As
hydrogen diffuses, it will change the bonding configuration of
the silicon network. Figure 4.3 shows IR spectra of sample #3
before and after annealing. The peaks at -2100 cm~l
represented the Si-H2 and Si-H3, or surface-like Si-H (eg. on
void surface), or 0-Si-H bonds had decreased after annealing.
Another beneficial result of hydrogen diffusion is that it
will compensate many in-gap defect states and change the
conducting mechanism from localized hopping among gap states
to extended state transport. Figure 4.4 shows the ESR of
sample #3 before and after annealing. The as-deposited
dangling bond density ng = 2.0x10-^ cm"^ ^as very high. After
annealing at 300°C for 9 hours, it decreased so much that the
ESR was undetectable. However the upper limit was estimated to
be at -1.0x10-'-° cm"^. In other words, the density decreased by
more than two orders of magnitude after annealing.
55
as deposited
3SB0 3S00 2500 !59S l o a a 5SB
\Vavenumber (cm')
Fig. 4.3 IR spectra showed the decreased intensity at -2010
cm"^ after annealing
56
5.0
4.0 -
10
0.0
<
H.O
A /
as deposit ns=2.0xl0l8 /cm3
^
\l . . . .
3.23 3.31 3.33
0.3
0.2 -I
0.0
-0.1
-0.2
I
-0.3 -
-0.4 -
-0.5 —1
-0.5 -
-0.7
3.35 3.37 3.33
- 'A jt, 1
"- \Af H r, / \
300 oc 9 hrs
ng < 1.0x10^6 /cm3
\j y \ u 1 / I
1 ' 1 1 ' ' ' , lU J I r 1
\ / \ f \ . A ! '•i .1 k.
im 1' 1.'
\ ,1
3.23 3.31 3.33 3.35 3.37
Magnetic Field (K Gauss) 3.33
Fig. 4.4 ESR raeasurements showed the dramatically decreased
dangling bonds after annealing
57
Unlike a-Si:H deposited by glow discharge (GD)
decomposition of silane, which normally has a strong SWE, the
samples prepared by rf sputtering and treated by annealing
displayed enhanced light soaking stability. The samples with
high Ch and low Pj-f showed a weak S'WE (see Figure 4.5 and
compare with Figure 1.4), while the samples with low Cjj and
high P>-f showed no SWE. It is not clear what aspect of the
deposition process and the annealing is responsible for the
superior stability against light-induced degradation in these
samples. Small angle x-ray scattering studies^S found that the
microvoid content increased with increasing hydrogen content
in rf sputtered a-Si:H films, and the SWE is generally
believed to result from light induced defects which reduce
both photo- and dark-conductivity. In that sense the enhanced
light soaking stability might be explained as the high initial
defect density indicated by which made the light
induced defect density relatively small. Generally, device
quality GD a-Si:H with similar high hydrogen content (-10 at.
%) show strong SWE and ~ 10^.
In conclusion, the electronic and optical properties of
rf sputtered s-Si:H sample are strongly improved after
annealing at 250 - 300°C; can be as high as lO'^, which
is close to GD a-Si:H. As expected, the SWE increased with
increasing cr^h/cJj. The sample with the highest showed
58
10 -3 I I I r I I "i—r
«
o
t3! c o o
10 -4
a 10 o
-5
^ 10"^
10 - 7
10
10
- 8
h 1 n -9 i \ j
- 1 0 L 1 1 1 ]
Illumination
j l
I
=3 —t q 1 1
~i
I I I I ! ' I ] n 9 4 jl
Time (iiours)
a
Fig. 4.5 Weaker Staebler-Wronski effect (coir.pare with Fig.
1 a •\
59
much weaker SWE, while low cTpj^/a,^ samples deposited at high rf
power and low H partial pressure showed no SWE at all.
B. Microvoids and K Content
Five a-Si:H were prepared for SAXS studies by rf
sputtering of a Si target onto a nominally unheated grounded
substrate located 1.5" below the target. The Ar and H2 partial
pressures were 10 and 0.5 mT respectively, and the rf power
varied from 200 to 600 W. The effective temperature of the
film during deposition was estimated to be -150°C.^5 For each
sample, three different substrates were used. The first was a
standard crystalline silicon wafer, which was used for IR
characterization, the second was Corning 7059 glass for
optical gap and photoconductivity characterizations, and the
third was a 10 jim thick iron-free aluminum foil for SAXS
measurements. Samples were characterized by IR and SAXS after
each annealing step, which was carried out in evacuated pyrex
tubes and lasted 6 hours.
The properties of the as-deposited samples are presented
in Table 4.2. As clearly seen, sample #1 and #2 had the higher
hydrogen content and larger microvoid volume fraction Vf;
in samples #3-#5 Cu was -10% and Vf was -0.5 vol.%. These
values are close to device-quality GD films. All samples
were about 2 um thick. After the first annealing step at
250 c cv of sstp.'oxs —!l <5sc2r05ssd, 22.% 2.3%
60
Table 4.2 As-deposited properties of the rf sputter-deposited
a-Si:H films, including H content Cjj and Si-H2 bond
density 0^2? SAXS determined microvoid content Vf,
and the mass density p determined by flotation
measurements
Sample # rf power
(W)
thicknes
s
(lim)
(at.%)
^H2
(at.%)
Vf
(vol.%)
P
(g/cm3)
1 200 2 . 4 5 2 0 . 8 3 . 4 1 . 7 * 2 . 1 8
2 3 0 0 2 . 4 0 1 9 . 0 -10* 2 . 0 4
3 4 0 0 2 . 7 0 1 1 . 2 - 0 0 . 5 4 2 . 2 2
4 500 2 . 0 0 1 0 . 2 - 0 0 . 5 1 2 . 2 5
5 600 1 . 9 5 9 . 5 - 0 0 . 4 7 2 . 2 4
* The actual void fraction is less, because the tilt-
dependence of the scattering intensity causes an overestimate
of Vf .
61
while Vf increased from 2 to 2.25 vol.%. The other samples
showed similar slight changes except sample #2 which exhibited
strong post-depositional oxidation after 75 days in air (as
deterinined by IR absorption) and was not annealed. Further
annealing from 250°C to 310°C produced basically no changes in
either Cjj or Vf (see Figure 4.6). Other samples prepared under
similar conditions and annealed in this temperature range
showed sharply reduced dangling bond densities and C7ph/C7(j
ratios of -10^.^^ After annealed at 350, 380, 400, and 430°C,
however, Cjj decreased significantly. Further annealing at
43 0°C for 3 0 more hours further decreased and sharply
increased Vf. The drop of hydrogen content when annealed at
above 330°C was consistent with others studies^^ that Si-H
bonds begin to break and the hydrogen evolves from the film
above 3 3 0°C.
If microvoids are simply approximate to be spherical,
then the SAXS also yields the void size distribution as
described in Eq. (3.15). Figure 4.7 shows the size
distributions of sample #l and #4 before and after annealing
process. They yield several features which should be
addressed: (i) They ail indicate a significant content of
relatively large voids, regardless of the overall morphology
of the films, (ii) While most of the voids in unannealed
sample #4 are -1 nm across, most of those of unannealed sample
#1 are 1 - 3 nm in diameter. This observation strongly
62
25
S o m p l e
20
o
c
o c
25
o 2 0
CJ •a o £
o
o
Z i
o o Scmp
o
150 200 250 300 350 400 ' .0 70
Annecl lempcrcture ( C) Time (h)
Fig. 4.6 The integrated SAXS intensity or the microvoids,
and the hydrogen content after each annealing step
63
Fig. 4.7
Sample 1
as —deposifed ^ 430°C X 35 h ^ 40
Sample 4
as —deposited 4-30 C X 35 h
10^
D i a m e t e r (n m )
Tne distribution of void sizes in sample #i and #4
before and after annealing
64
suggests that the high overall inicrovoid content of sample #1
is clearly due to the large size of microvoids rather than to
more numerous small voids, (iii) After annealing, the
distribution of void sizes are wider in both samples, but the
large-scale density fluctuations are clearly larger in sample
#1 than in #4.
Three major conclusions resulting from the measurements
described above are clear: (a) They are consistent with the
well-known correlation between Cm, 0^2r and Vf: The initial
Si-bonded H content of Sample #1 and #2 is high (-20 at. %),
and it indeed contains a significant dihydride concentration.
The tilt angle-dependence of the SAXS clearly indicates that
the microstructures are columnar-like. The SiH2, SiH3, and
microvoid density of all of the other samples, which contain
9-11 at. % Si-bonded H, is very low. Both SAXS intensity and
Cu for the noncolumnar samples (#3-#5) did not change during
prolonged annealing (beyond 6 hrs) at 430°C. This behavior is
believed to result from migration of m.ost of the Si-bonded H
to the internal surface of that film, recombination to
molecular H2, and its escape through the largely
interconnected voids. Most of the remaining hydrogen of that
film was then probably bonded to the void surface in isolated
sites which are not adjacent to a neighboring H atom, (b)
Annealing at 350°C and above sharply increased the integrated
SAXS intensity,, or increased the average void size, and
65
decreased the Si-bonded H content C^. This behavior was
apparently due to hydrogen migration from the Si network to
microvoid surface, recombination of adjacent H atoms at the
surfaces, and accumulation of molecular K2 in isolated voids.
The decrease in Si-bonded H content then increased the density
of the Si network. Its reconfiguration and probable high
pressure of the trapped molecular hydrogen then increased the
average void size. It should be noted that decreasing density
of molecular H2 through evolution may also results in the drop
of SAXS detected Vf, however, the magnitude of the changes in
the SAXS may exclude such a possibility, (c) Annealing at
430°C resulted in the formation of a sufficient volume
fraction of microcrystalline silicon (|j.c-Si) domains to be
detected by Raman and XRD. Although the amorphous phase was
still dominant after 12 hrs at 430°C, it was reduced to -50%
after 3 6 hrs at the same temperature, and the columnar
structure was almost eliminated. This behavior is not
surprising in view of the large-scale crystallization
processes occurring when annealing at such a temperature
range.
C. Anomalous Hydrogen Diffusion
The time dependent diffusion constant of hydrogen in a-
Si:H was found to obey a power-law given
66
D(t) = Doo(®t)-a 4.1
where a is the dispersion parameter and co and DQQ are
constants. In the multiple-trapping (MT) model which has been
frequently invoked to describe hydrogen motion in a-Si:Hl2,14
it is assumed that the density of hydrogen sites is
exponentially distributed in trapping energy Ef This
exponential distribution of trapping energies leads to
a = 1 - T/Tq 4.2
where T is the annealing temperature and RTq is the
distribution width of trapping energies In this
multiple trapping picture of hydrogen motion, the hydrogen
will be trapped into the deeper and deeper sites as it
migrates. The longer it diffuses, the deeper it will be
parameter a should always be greater than zero. Indeed, all
values of a reported to date were between 0 and 1. However,
the dependence of a on T varied widely among different
samples. In doped GD films, the results were consistent with
Eq. (4.2) but not sufficiently detailed to confirm In
undoped GD films deposited at high rf frequency a actually
strongly increased with T above 350°C.®2 In some rf sputtered
67
15.0 300° C
12.5
185 hrs 10.0
7.5
5.0 724 hrs. 50 hr^l 85 hrs.
2.5 30
330° C z)
<
CO 20
co 30 hrs> 72 hrs>\146 hrs. 293 hrs. _
25 363° C
20
-c r \
24 hrs> • i"v
48 hrs.
0 0.05 0.1 0.15 0.2 0.25
Diffusion D!st3.nc6
Fig. 4.8 Fitting of sample A between SIMS profiles (dots)
and complementary error function [Eq. (1.6)]
68
i i i i i i i i i i
3 <
CO cz 3 o o cd
00
- 1 I I I I I I 1 I
325° C
M
2, 4, 10. 20 hrs
343° C
~ \V\ 1.2, 2.4, 4.7 hrs
362° C
.A3.0 hrK7.2 hrs. 15.7 hrs. i i rs ! i i i
1.0 hrs;
0 0.05 0.1 0.15 0.2
uiTfusion Distance (ji)
0.25
Fig. 4.9 Fitting of sample B between SIMS profiles (dots)
and complementary error function [Eq. (1.6)]
69
384° C
1.2 hrs
20 -0.8 -0.4
8.8 hrs.-Z ) <
1.2 hrs. 4.7 hrs.
c
0 25 CD
1 20
405° C
C O
1.2, 4.8 hrs.
0.2 0.25 u.uo
Diffusion Distance (u)
Fig. 4.10 Fitting of sample B between SIMS profiles (dots)
and complementary error function [Eq. (1.6)]
70
samples used in this work, we found and reported^"^ a strong
deviation from the multiple trapping model, in which a was
negative i.e., the D(t) increases with t.
The films were deposited by rf sputtering at 550 W,
annealed, and monitored by IR and SIMS as described earlier.
Sample A was a 2.4 iim thick a-Si:H/a-Si: (H,D)/a-Si:H trilayer,
and each layer was deposited over 3 0 minutes- Sample B was a
1.65 jim thick a-Si: (H,D)/a-Si:H bilayer, grown at 30 minutes
per layer. The IR stretch band of both samples before and
after each annealing peaked at 2000 cm~^, indicating almost
exclusive bulklike Si-H bonding. Cj^, as determined from the
640 cm~l wagging mode vibration, generally decreased from -5
to -3 at. % over the annealing periods shown in Figures 4.8 -
4.10.
9(t) defined by Eq. (1.5) was obtained by fitting the
deuterium SIMS concentration profile c{x,t) to complimentary
error function as given by Eq. (1.6). As clearly seen from
Figures 4.8-4.10, the agreement between the SIMS profiles and
the complimentary error function was always reasonable, and
usually excellent.
Substituting Eq. (1.7) into Eq. (1.5), we get
o - Oq-C^ ^ 4.3
71
X 300°C, a=-0.47
V 330°C, a = -0.34 o 362°C, a = -0.04
r io
£ u
a:>
SAMPLE A
7 s
Annealing lime (sec)
Fig. 4.11 logiQ[0(t)] vs. log]_o[t] of sample A. Note that a
slope larger than 1 implies a < 0 [see Eq. (4.3)]
72
10 -10
- D 325°C, a = 0.31 -
x 343°C,a= 0.15 ~ + 36Z°C,a=-0.79 —
: o 384°C, a = -0.51 -
- V 405°C,a = 0.43 -
10"!'
10-'2
!q-'2 u_
i O -
SAMPLE B
i i i i i i i
!0"
"ieclip.g Time (sec)
Fig. 4.12 logi_o[0(t)] vs. logi_n[t] of sample B. Note that a
slope larger than 1 implies a < 0 [see Eq. (4.3)]
73
Therefore, the slope of logio[0(t)] vs log3^Q[t] is 1-a. Figure
4.11 shows such a plot for sample A annealed at 300, 330, and
362°C. Remarkably, the value of a is -0.47 at 300°C but
increases to -0 at 362°C. Figure 4.12 exhibits a similar plot
of sample B in the temperature range of 325 - 405°C. From 325
to 362°C a decreases from 0.31 to -0.79, but then increases to
0.43 at 405OC.
Both samples A and B exhibited (i) negative values of a
and (ii) an a which increased with T at high annealing
temperatures. All of these results are contradictory to the MT
model, and attributed to the complex relaxation processes in
which the distribution of H site energies is probably
modified. If the shallower band narrows during annealing at
300 < T < 350°C, and the deep band rises toward the transport
state (see Figure 4.13), then Dj^(t) should increase with time
and a should be negative. Such a behavior would be expected if
the shallow band is due to interstitial sites similar to those
of H in crystalline Silicon, since the activation energy of
atomic H diffusion in crystalline silicon is -0.5 eV.^^
Although such processes may entail a rise in the free energy
Fjj of the H atoms, the total free energy of the H and Si atoms
may decrease. Roorda et al.^® indeed found that Si amorphized
by ion implantation undergoes exothermic relaxation beginning
at temperatures as low as 110°C, v/hen heated as rapidly as
4Q°C/ir.in. The structural relaxation occurring during several
74
Transport level
>
LiT
U)
o c LU U) c •q. q. cc
h-
x
l^ciioiiy vj 5 vjllcto
Fig. 4.13 The suggested distribution of H trapping energies
in a-Si:H. Hpj is the chemical potential of H atoms.
See Ref. 64
75
hours at T > 300°C should therefore be considerable and may-
affect the shallow H sites and the position of the deep H band
in the aforementioned manner.
As mentioned earlier, the SAXS study of a-Si:H^2
indicated that microvoid content Vf steadily increases when
annealing at 250 < T < 430°C. As the deepest H-trapping sites
are believed to be mono Si-H bonds on internal microvoid
surfaces, we conclude that the strong increase of a with T is
due to an increasing microvoid content. The concentration of
hydrogen in shallow sites thus decreases as an increasing
number becomes deeply trapped on the microvoid surfaces.
76
references
1. W. E. Spear and P. G. LeComber, Solid State Commun. 17,
1193 (1975).
2. G. C. Chittick, J. H- Alexander, and H. F. Sterling, J.
Electrochem. Sco. 116, 77 (1969)
3. A. J. Lewis, G. A. N. Connell, W. Paul, J. R. Pawlik, and
R. J. Temkin in Tetrahedrallv Bonded Amorphous
Semiconductors. edited by M. H. Broidsky, S. Kirkpatrick,
and D. weaire, AIP Conference Proceedings, 20, 27 (1976).
4. W. Paul, A. J. Lewis, G. A. N. Connell, T. D. Moustakas,
Solid State Commun. 20, 969 (1976).
5. R. J. Temkin, W. Paul, and G. A. N. Connell, Adv. Phys. 22,
581 (1973).
6. A. Matsuda, M. Matsumura, K. Nagiakgaqwa, T. Imura, H.
Yamanoto, S. Yamusaki, H. Okushi, S. lizima, and K.
Tananaka, in Tetrahedrallv Bonded Amorphous Semiconductors,
edited by R. A. Street, D. K. Biegelson, and J. C. Knights,
American Institute of Physics, New York, 1981, p. 192.
7. S. Zhao and S. Kunklinger, Solar Energy Mater. 6, 22 3
(1982).
8. D. L. Staebler and C. R, Wronski, Appl. Phys. Lett. 31, 292
(1977).
9. D. L. Staebler and C. R. Wronski, J. Appl. Phys. 51, 3262
(1980).
77
10. M. Stutzmann, W. B. Jackson, and C. C. Tsai, Phys. Rev. B
32, 6062 (1985).
11. R. A. Street, C. C. Tsai, J. Kakalios, and W. B. Jackson,
Phil. Mag. B 56, 305 (1987)
12. J. Crank, The Mathematics of Diffusion. Clarendon, Oxford,
England, 1975, Chap. 2.
13. J. Kakalios, R. A. Street, and W. B. Jackson, Phys. Rev.
Lett. 59, 1037 (1987).
14. W- B. Jackson, Phys. Rev. B 38, 3595 (1988).
15. n . B. Jackson and J. Kakalios, Phys. Rev. B 37, 1020
(1988).
16. J. Shinar, R. Shinar, S. Mitra, and J. Y. Kim, Phys. Rev.
Lett. 62 2001 (1989).
17. W. B. Jackson, Phys. Rev. B 41, 1059 (1990).
18. W. B. Jackson and M. D. Moyer, Phys. Rev. B 36, 6217
(1987).
19. M. Albers, Ph.D. Thesis, Iowa State University, (1987)
unpublished.
20. J. Tauc, R. Grigorovicci, and A. Vancu, in Amorphous and
Liquid Semiconductors, edited by J. Tauc, Plenum, New York,
1974.
21. F. Urbach, Phys. Rev. 92, 1324 (1953).
22. M. H. Brodsky, M. Cardona, and J. J. Cuomo, Phys. Rev. B
16, 3556 (1977).
78
23. H. R. Shanks, C. J. Fang, L. Ley, M. Cardona, F. J. Demond,
and S. Kablitzer, Phys. Status Solidi (b), 100, 43 (1980).
24. C. J. Fang, K. J. Gruntz, L. Ley, M. Cardona, F. J. Demond,
G. Muller, and S. Kalbitzer, J- Non-Cryst. Solids 35 & 36,
255 (1980).
25. M. Cardona, Phys. Status Solidi (b), 118, 463 (1983).
26. H. R. Shanks, F. R. Jeffrey, and M. B. Lowry, J. de
Physique 42, C4-773 (1981).
27. H. Wagner and W. Beyer, Solid State Commun. 48, 585 (1983).
2S. C. P. Poole, Jr., in Electron Spin Resonance. 2nd Ed.,
Wiley, New Yo-^k, 1983, p. 409.
29. F. M. Smits, Bell Sys. Tech. J. 710 (1958).
30. The SIMS measurements were performed by Dr. Ruth Shinar at
Microelectronics Research Center, Iowa State University,
Ames, Iowa 50011.
31. J. Baum, K. K. Gleason, A. Pines, A. N. Garroway and J. A.
Reimer, Phys. Rev. Lett. 56, 1377 (1986) .
32. Y. J. Chabal and C. K. N. Patel, Rev. Mod. Phys. 59, 835
(1987).
33. H. Wagner and W. Beyer, Solid State Commun. 48, 585 (1983).
34. A. K. Mahan, P. Raboisson, D. L. Williamson and R. Tsu,
Sol. Cells 21, 117 (1987).
35. A. H. Mahan, P. Raboisson and R. Tsu, Appl. Phys. Lett. 50,
335 (1987).
79
36. A. H. Mahan, P. Menna and R. Tsu, Appl. Phys. Lett. 51,
1167 (1987).
37. W. Beyer, H. Wagner and H. Mell, MRS Symp. 49, 189 (1985).
38. V. G. Bhide, R. O. Dusane, S. V. Rajarshi, A. D. Shaligrain
and S. K. David, J. Appl. Phys. 62, 108 (1987).
39. Y. J. He, M. Hasegawa, R. Lee, S. Berko, D. Adler and A.-L.
Jung, Phys. Rev. B 33, 5924 (1986).
40. S. C. Moss and J. F. Graczyk, Phys. Rev. Lett. 23, 1167
(1969) .
XT T T.T -r Oxiov J./V O.Xi^ ri* X'OU.X/ U * O OOX
(1972) .
42. G. S. Cargill, Phys. Rev. Lett. 28, 1372 (1972).
43. R. J. Temkin, W. Paul and G. A. N. Connell, Adv. Phys. 22,
581 (1973).
44. N. J. Shevchik and W. Paul, J. Non-Cryst. Solids 16, 55
(1974) .
45. P. D'Antonio and J. K. Konnert, Phys, Rev. Lett. 43, 1161
/ r \ r s \ v x 3 / .
46. A. <J. Leadbetter, A. A. M. Rashid, R. I»I. Richardson, A. F.
Wright and J. C. Knights, Solid State Commun. 33, 97 3
(1980) .
47. P. D'Antonio and J. H. Konnert, AIP Conf. Proc. 73, 117
(1981).
48. J. F. Graczyk, Phys. State Sol. (a) 55, 231 (1979).
80
49. R. Vanderhaghen, B, Chaurand and B. Drevillon, in P. Pinard
and S. Kalbitzer (eds.), Polv-microcrvst.alline and
Amorphous Semiconductors. Editions de Phys., Les Ulis,
France, 1934, p. 661.
50. R. Vanderhaghen, B. Chaurand and B. Drevillon, Thin Solid
Films 124, 293 (1985).
51. T. A. Postol, C. M. Falco, R. T. Kampwirth, I. K. Schuller
and W. B. Yelon, Phys. Rev. Lett. 45, 648 (1980).
52. A. J. Leadbetter, A. A. M. rashid, N. Colenutt, A. F.
Wright and J. C. Knights, Solid State Commun., 38, 957
(1981).
53. R. Bellissent, A. Chenevas-Paule and M. Roth, J. Non-Cryst.
Solids 59/60, 229 (1983),
54. SAXS measurements were performed by Mr. Yen Chen at
Department of Physics, Colorado School of Mines, Golden, CO
80401.
55. D. L. Williamson, A. H. Mahan, B. P. Nelson and R. S.
Crandall, Appl. Phys. Lett. 55, 783 (1989).
56. A- Guinier and G. Fournet, Small-Angle Scattering of X-
ravs. John Wiley, New York, 1955, p. 30.
57. G. Porod, in Small Angle X-rav Scattering, edited by O.
Glatter and O. Kratky, Academic Press, New York, 1982, p.
17.
58. H. Jia, J. Shinar, Y, Chen, and D. L. Williamson, Mat. Res.
Soc. Proc= 258, 281 (1992).
81
59. M. L. Albers, J. Shinar, and H. R. Shanks, J. Appl. Phys.
64, 1859 (1988).
60. A. H. Mahan, D. L. Williamson, B. P. Nelson and R. S.
Crandall, Sol. Cells 27, 455 (1989).
61. H. Jia, J. Shinar, and H. S. Wu, Bull. Am. Phys. Soc. 37,
128 (1992) .
62. D. K. Biegelsen, R. A. Street, C. C. Tsai, and J. C.
Knights, Phys. Rev. B 20, 4839 (1979).
63. X.-M. Tang, J. Weber, Y. Bear, and F. Finger, Phys. Rev. B
I ^ 4 t \ Zf \J ) •
64. R. Shinar, J. Shinar, H. Jia, and X.-L. Wu, Phys. Rev. B
47, 9361 (1993).
65. S. J. Pearton, J. W. Corbett, and T. S. Shi, Appl. Phys. A
43, 153 (1987)
66. S. Roorda, S. Doom, W. C. Sinke, P. M. L. O. Scholte, and
E. van Loenen, Phys. Rev. Lett. 62, 1880 (1989).
82
farr xwo.
fabrication and characterization
cvd diamond films and devices
83
I, INTRODUCTION
A. Historical Background
Diamond is one of the most attractive materials, as it
has a unique combination of excellent mechanical, physical,
and chemical properties.^ The well known properties of diamond
are summarized and compared with silicon in Table 1.1. Its
4000^0 melting point and its resistance to all chemical acids
make it an ideal material for harsh environments. The wide
band gap of diamond can be used in far-ultraviolet detectors
to eliminate the visible and infrared background which
normally appear in silicon based detectors. It is a highly
electrically insulating material but can be doped both p and n
type, and its superb thermal conductivity render it an
excellent heat sink which, if applied in integrated circuits,
will enable a sharp increase in the density of integrated
circuits.
The industrial demand for diamond was inhibited by its
prohibitive cost or unavailability until the l950s2/2 when
General Electric successfully developed a high-pressure high-
temperature (HPKT) method to synthesize it. Today 30 - 40 tons
of HPKT synthetic diamond are produced worldwide each year,
and all of it is used for cutting, grinding, and polishing
tools. However many potential applications of diamond require
84
Table 1.1 Comparison of semiconductor properties between
diamond and silicon (from Ref. 1)
Properties Diamond Silicon
Lattice constant (nmi (A)) 0.3567 (3.557) 0.5430 (5.430) Thermal expansion (xlO'V''C)1.1 2.6 Density (g/cm^) 3.515 2.328 Melting point CC) 4000' 1420 Band gap (eV) 5.45 1.1 Saturated electron velocity
(x 10' crn/s) 2.7 1.0 Carrier mobility (cm'7(V • s))
Electron 2200 1500 Hole 1600 600
Breakdown (xio^ V/cm) 100 3 Dielectric constant 5.5 11.8 > - , . » . , » c
«-( OQ 1 Qin/I7> ' f\ } . . . w w w J I J
i n*3 1 \J
-< r\3 i \J
Thermal conductivity (W/(cm • K)) 20 1.5
Absorption edge (/xm) v.d. 1.4 Refractive index 2.42 3.5 Hardness (kg/mm') 10000 1000 Johnson figure of m.erit
( x lO^^ (W-0 ) / s2 ) 73 856 Q O ^
Keyes ligure oi merit ( x lO -VV / ( cm-s - "C ) ) 444 1 « » W . V-/
85
thin films of coating which can not be produced from either
natural or HPHT synthetic diamond.
The phase diagram shown in Figure 1.1 clearly indicates
that diamond is stable only in the high pressure high
temperature region as obtained in the GE process. Yet that is
not the only way that diamond can be synthesized. If the
activation energy between stable and metastable states is high
enough, it serves as a barrier to interconversion, and
metastable material can be formed under kinetically controlled
conditions similar to numerous other reactions. Hence a
parallel effort directed toward the growth of diamond at low
pressures where it is the metastable state was explored
several decades ago.
The first successful attempt was reported by W. G.
Eversole^ who initiated the work in 1949 and achieved growth
on diam.ond seed at the end of 1952 by using decomposition of a
carbon-containing gas. In Eversole's work, the diamond and
graphite were first codeposited in a seed crystal, then the
substrate and deposit were transferred into a separate
autoclave. The graphite was removed by hydrogen etching at
temperatures T > 1000°C. The process was repeated many times
to achieve the desired thickness. Angus and co-workers
confirmed Eversole's results.They also studied the rate of
diamond and graphite growth in a CH4 - H2 gas mixture and v/ere
the first to report on the preferential etching of graphite
86
Pressure, kbar 400
?nn . /\j\j t
200
100
0
Shock wave synthesis.
Liquid carbon
a ' 1 -
Diamond & metastable graphite
Diamond \ Catalytic high-pressure, high-temperature synthesis
High-pressure, high-temperature synthesis
Chemicai vapor u e pO S1 liO n
! !
Graphite & g metastable M diamond ^
u 1000 2000 3000
Temperature, °C
conn ou vu
Fig. 1.1 Carbon phase diagram showing the stable diamond
phase at high pressure and temperature and
metastable region at low pressure
87
vs. diamond by atomic hydrogen and to demonstrate the
deposition of boron-doped semiconducting diamond.
The worldwide interest in low pressure chemical vapor
deposition (CVB) of diamond started in 1982 after N. Setaka,
Y. Sato, M. Kamo, and S. Matsumoto in Japan published a series
of papers describing in detail the techniques for synthesizing
diamond at rates of several |im per hour from microwave or DC
discharges, or from gases decomposed by a hot filament.
The processes produced individual faceted crystals without
using diamond seed crystals. Since then the CVD of diamond
from hydrocarbon precursors has been extensively studied and
become the most successful technique for metastable diamond
growth.
B. CVD of Diamond Films
Thin diamond films have been successfully grown at low
pressure by a large variety of energetically assisted CVD
processes. These can be basically divided into two groups:
thermally assisted CVD, and plasma assisted CVD (PACVD). They
all share a common feature: the carbon-containing gas is
decomposed into atom.ic species either thermally or by a
plasma.
Figure 1.2 shows a hot filament assisted CVD system,
which is typical of the thermally assisted process. The
filament, which inay be W. Ta.. Mo_. or Re, is heated un to
88
AC power
SUDDIV
t
Gas inlet
Heater
CH Gas exit
Substrate
Fig. 1.2 Schematic of hot filament assisted CVD system
89
2500°C. The CH4 gas is heavily diluted with H2 at a ratio of
-1/100; -5% of the mixture is then dissociated into atoms or
radical fragments when it passes through the hot filament. The
substrate which is about 1 cm under the filament is kept at
70D-1000°C. The advantage of hot wire CVD is its ease of
operation. The disadvantage is that the film may be
contaminated by emission of metal atoms from the hot filament.
Figure 1.3 shows a DC plasma assisted CVD. A stabilized
DC plasma produced growth rates up to -20 jam per hour at the
anode. However, only graphite deposited on the cathode. The
typical applied voltage and current density are 1 kV and 4
A/cm^ respectively.
The DC plasma jet technique (Figure 1.4) focuses the
diamond coating on a relatively small (5x5 mm^) area, but the
plasma can be scanned over a large area to produce a uniform
deposition. This technique uses a high temperature plasma in
which nearly all molecules are completely dissociated. Growth
rates of more 500 jim per hour have been achieved by this
method.
Microwave plasma CVD offers another way to deposit
diamond films. Figure 1.5 shows a bell jar reactor for this
technique. The microwave radiation is guided to the top of the
reactor by a rectangular metal wave guide and picked up by an
sntGnns tlist p^rotiruciss into ths ci^rculsr*. Tiis iTiixtiiirs cz
m o 3 o p >-i •» c- ^ • c- ^ •— W - » • W . • W. • • jj' S_> C >>.A. C% * « .4- Alt Ck. O A 1 ±. CX
90
Water Gas exit
Substrati
holder
DC
I I — r r
Gas
inlet CH
Water
Fig. 1.3 Schematic of a DC plasm.a assisted CVD system
91
dc source
i 1 { ! Argon & hydrogen
Cooling
Methane
Plasma jet
• vw X caic
U*«-A.>S.,JUWK.A *—Al*«^
! 1 •
i T » • aako/^ w'wv./mti^ riuici
Fig. 1.4 Schematic of a DC plasma jet system
92
AnlGnns
Microwaves
Rectangular metai waveguide
Circular, transparent metal waveguide
Silica bell jar
Heater
Ball-shaped plasma -
Subslrale
MCIMSHB & hydrogen
i
m
\ r ^ \
S T o D u m p
r—1 m 3
\\\\\\\\\\\X\\\\\\\\\\\\\\\\\ ^
Fig. 1.5 Schematic of a bell jar microwave reactor system
93
FLOWMETERS TORCH
o2
c2h2
SUBSTRATE
PYROMETER
f1 I
WATER
X Cu MOUN
INNER CONb
ACETYLENE PEA 1HtH
OUTER t-LAMt
Fig. 1.6 Schematic of an cxy-acetylene torch CVD sysrsm
94
plasma above the substrate. This technique can deposit diamond
films on substrates up to 8 cm in diameter.
Another noteworthy technique is the combustion flame
shown in Figure 1.6, which uses an oxygen-acetylene torch and
a water cooled substrate. It can deposit high quality diamond
films in an ambient atmosphere, and the growth rate can be as
high as 180 [im per hour.l^"^^
C. Growth Mechanism of CVD Diamond
As discussed earlier and shown in the phase diagram
(Figure 1.1), graphite is the stable form of carbon under the
conditions used in CVD of diamond. Why is it then possible to
grow diamond at less than atmospheric pressure at temperatures
750° < T < 1100°C?
Although an established mechanistic answer to this
question is not yet available and is still the subject of
current research, one model consistent with experimental
results was developed by Spear and coworkers. 0 ? 21
mechanism is based on the fact that growth occurs at the gas-
solid interface of the carbon-hydrogen system. The vapor-
growth process does not involve elemental carbon only, but
also hydrogen. A diamond surface saturated v.'ith sp^ C-K bonds
is more stable than a diamond surface free of hydrogen. Once a
surface carbon is covered by another diamond growth layer.
95
DIAMOND
GRAPHITE
structure of diamond and graphite. The hydrogen
atoms depict their role in stabilizing the diamond
surface
96
then that covered carbon possessing four sp^ C-C bonds is
metastable with respect to graphite.
Lander and Morrison^^ were the first to hypothesize that
hydrogen can stabilize a diamond surface by forming sp^ C-K
bonds with surface carbon. Without the hydrogen maintaining
the sp^ character of these surface carbons, it is easy to see
that the diamond surface will collapse into the more stable
planar graphite structure (Figure 1.7) during the growth
process.
Hydrogen is also responsible for other important
mechanisms during diamond growth. As proposed by Frenklach and
Spear23^ the surface bonded H is first removed by atomic H
from the gas phase to form an activated carbon radical and H2
c c \ \
H- + C-H J- H2 + C- 1-1
/ / c c
then this surface-activated carbon radical acts as a site for
adding more carbons to the structure by reacting with
acetylene or other carbon-hydrogen species in the gas or
plasma, e.g.,
C C H H \ \ i ' N \ I / C- + K-C=C-K > C-C=C* --2
/ / c c
97
the C=C graphite-like double bond can be reconstructed into
sp2 bonds by H2•
C C C C
\ / \ / C=C + H-H > H-C-C-H ^-3
/ \ / \ c c c c
Deryagin and Fedoseev^^ proposed that a "super
equilibrium" concentration of atomic hydrogen at the growth
surface is responsible for the major reduction in graphite
codeposition. They argued that atomic hydrogen behaves like a
"solvent" for graphite. In fact, numerous studies24-26 show
that at T > 8 00°C atomic hydrogen etches graphite at rate
almost 500 time faster than it etches diamond. That results in
the net growth of diamond under metastable conditions.
D. Scope or Tnis worx
'T'1 0 vork began with the constructiori of a new hot
filament assisted CVD system for diam.ond deposition by using
minimum, resources of equipment that already existed in the
laboratory. After the successful deposition of films, the goal
was expanded into (i) studies of native defects in thin
diamond films, and (ii) fabrication of light emitting diodes
(LEDs) from both intrinsic and doped films. The major
techniques used in this study were x-ray diffraction (XRD) to
98
characterize the diamond structure, scanning electron
microscope (SEM) to characterize the morphology of the films,
and ESR spectrometry to study their native defects.
99
II. SAMPLE FABRICATION
A. Deposition System
All samples used in this work were prepared by using a
home built hot filament assisted CVD system. A schematic of
the deposition system is shown in Figure 2.1. Using a
mechanical pump, the deposition chamber was routinely
evacuated to a base pressure of -3xlO~2 Torr prior to each
deposition. The 0.5 mm diameter tungsten filament was heated
to about 2000^C by an AC power supply; the temperature of the
filament was monitored with an optical pyrometer. The
substrate was about 1 cm underneath the filament and kept at
about 800-900°C by radiation from the filament. The 0.5%/99.5%
CH4/H2 gas mixture was introduced into the chamber at a flow
rate of -200 standard cubic centimeters per minute (seem). The
deposition pressure of 50 Torr was obtained by closing the
main valve connecting the chamber and the mechanical pump, and
adjusting the micrometer valve in the bypass line.
B. Intrinsic and Doped Sample Preparation
The diamond films were deposited on 1.5x3 cm^ n or p type
(100) Si wafers. It is well known that diamond nucleation is
greatly enhanced and uniform polycrystalline diamond films are
obtained if the Si substrate is scratched with diamond powder
before deposition.The Si wafers were thus usually
100
/
TUHGSTEN FIIiMEKT
SUBSTRATE
ttrttt^ir/*v^/nrrn^ r * iniii\rika^uuci.r. —p a c?3
TO PUMP f-
\
I i I !
GLASS JAR
< chd &: k2
Fig. 2.1 Scneraatic of hot filaraent assisted CVD system
101
pretreated before deposition by either polishing or
ultrasonically bathing the wafers with 0.25 |iiti diamond paste
for about 30 minutes, followed by ultrasonic cleaning in
deionized water, acetone, and methanol. Figure 2.2 shows an
SEM image of the scratched surface.
Table 2.1 Typical parameters for CVD of diamond films
Parameter Range
CH4 (%) 0.5 - 2 . 0
h2 (%) 98.0 - 99 . 5
Total flow rate (seem) 180 - 220
Deposition pressure (Torr) 40 - 60
Filament temperature (°C) 1900 - 2200
Substrate temperature (°C) 800 -• 900
r>i ^ ^ n j-'u ^ ^ ^ z i. i ^ — . - _ ^ ^ 1 * A. ^ X J. w i. w j_ wi ic; V.A ^ o u. a. V-/1X w tz:
noted that the tungsten filament became carbonized. As a
result, the resistance of the filament increased and the power
supplied to filament had to be increased accordingly to keep
the temperature of the filament at about 2000°C. The whole
deposition process normally lasted more than 3 hours depending
on the desired thickness of the film. The typical growth rate
of the film was between 0.5 - 1.5 }j.m/hr depending on the
CH4/H2 ratio.
102
Fig. 2.2 SEM images of the scratched surfaces of the Si
wafer after polishing with 0.25 )im diamond paste
103
The extreme refractive properties of diamond inhibit
doping by means such as diffusion, etc. It has become a clear
advantage that CVD diamond can be doped during deposition
process by introducing the dopant in the reactant gas.27
Attempts to deposit n type films were also made in this work,
to improve the light emitting efficiency of diamond LEDs. The
dopant gas was N2 heavily diluted with H2.
C. Light Emitting Diode (LED) Structure
The technology for GaAs-based yellow-orange LEDs has long
been established. Many efforts have also been directed at
development of blue LEDs. Diamond is one of the candidates for
these devices, since it has a wide band gap of 5.4 eV, and a
high breakdown voltage of a few MV/cm. Although 5.4 eV
photons are clearly well within the UV range, an intragap
donor-acceptor transition yields blue electroluminescence
(EL). A LED structure as shown in Figure 2.3 was also
fabricated from diamond films to study the EL.
The diode fabrication process began first with the
deposition of a 3-10 [im thick diamond film, on silicon wafer.
The silicon substrate was then dissolved by HF to yield a
free-standing film. A thin metal film (A1 or Ga) of 0.2 ,um,.
and indium tin oxide (ITO), which is a transparent conductor,
were then deposited on each side of the free standing diamond
film by electron beam evaporation.
104
ITO
Fig. 2.3 Schematic structure of a thin diamond film LED
105
III. SAMPLE CHARACTERIZATION
A. Thickness Measurements
The thickness and surface roughness of the films was
measured using a Sloan Dektak profilometer as described in
Part One. SEM provided an additional method to determine the
thickness and morphology of the films by using a broken
section of the sample, as shown in Figure 3.1 in which the
thickness of the film was determined to be -3 0 jam.
B. Scanning Electron Microscopy
Samples were usually studied with an optical microscope
(Olympus Vanox) to monitor surface uniformity and morphology
on an optical scale. When deemed necessary, SEM images were o
also obtained. The SEM has a resolution of -250 A, and a
depth of field 300 times more than that of the optical
microscope, all of which result in photographs of dramatic
three-dimensional quality.
Prior to each SEM measuremenr, a thin layer of gold was
DC sputtered on the diamond film to prevent the charge
accumulation which can interfere with the scanning electrons
and cause smear output image. Figure 3.2 shows typical SEM
pictures of diamond films taken with a Cambridge Stereoscan
200 SEM. In general, high quality diamond films exhibit well
106
Fig. 3.1 SEM photograph of CVD diamond film,
of the film is determined to be -30
picture
the thickness
um from the
107
108801
Fig. 3.2 Typical SEM images of CVD diaip.ond filir.s
108
faceted crystal structure with (001) or (111) planes. Figure
3.3 illustrates the change from cubic to octahedral diamond.28
The diamond films were also characterized by powder x-ray
diffraction (XRD). They were directly mounted on a microscope
slide. The XRD spectra were obtained by using a microcomputer-
controlled Rigaku diffractometer equipped with a cooper target
and a diffracted beam graphite monochromator for Cu Ka
radiation with a step scan rate of 0.01 degree/second over the
20 angular range of 20-95 degrees.
The basis of XRD is Bragg's law^^ which describes the
condition for constructive interference for x-rays scattered
from atomic planes of a crystal:
Here d is the interplanar spacing of the crystal, 20 is the
angle of the diffracted beam, and X is the wavelength of the
incident x-rays as shown in Figure 3.4. The distance between
the (hkl) planes of a crystal is given by
C. X-ray Diffraction Measurements
2dsin0 = X 3.1
3 . 2
109
Cubo-octohedrons Cube
Octohedron
Fig. 3.3 Polyhedrals of diamond
110
where a = 3.56 A is the lattice constant of diamond. For
face-centered cubic (fee) structures such as diamond, the
indices (hkl) have to be either all even or all odd in order
to yield a diffracted beam. In this work the incident beam was o
the Cu Ka line of wavelength ?l=1.5418 A. The expected XRD
pattern from diamond, as determined by Eq (3.1) and (3.2), are
shown in Table 3.1.
Figure 3.5 shows a typical XRD spectrum of a diamond
film. Three sharp peaks are identified as the diffraction from
(111) , (220) , and (311) planes. A broad peak at 20 = 70*^ is
due to the (400) Si substrate.
D. Electron Spin Resonance (ESR) Measurements
ESR was discussed earlier in Part One, except for the
case of a hyperfine interaction between the electron and a
nuclear spin.
In this case, the spin Hamiltonian is given by^*^
H = cSB•S + AS•I 3 .:
Where the first term represents the normal Zeeman splitting,
the second term represents the hyperfine coupling between the
electron spin S and nuclear spin I, and A is the hyperfine
coupling constant. Assuming gp3*3 is much larger than AS"I,
Ill
Table 3.1 Expected XRD patterns of polycrystalline diamond
films
(hkl) (111) (220) (311) (400)
20(°) 44.1 75 . 5 91.8 120. 0
Incident Beam
Detector
Fig. 3.4 Schematic diagram of XRD measurement, the detector
can be scanned at rate of 0.01 degree/second over
range of 20-95 degree
112
1250
1000
^ 750
ijo C Q> 500
250
I I I i i 1 i I i i i i i i i i i i i I i 1 i
(111)
u • J l .
I rrr m i—i—:—i—:—:—l—1_
(220)
11
J 11 i U 1 L
_1
J
(311)
I ' ' ' W1 i 1 I I
40 50 60 70 80
Diffract angle- 29 (
or* 1 ao c5 v-/
Fig. 3.5 Typical XRD spectrum of CVD diamond- The broad
pattern at 20 = 7 0° is due to the Si substrate
113
the energy levels obtained from first order
perturbation theory:
where M and m are the magnetic quantum numbers of the electron
and nucleus, respectively. The level structure and allowed
electron spin transitions between them are illustrated in
Figure 3.5 for the case S=l/2, 1=1/2.
The allowed ESR transitions are AM=±1, Am=0. Therefore,
the energy absorbed by the spin system when a microwave photon
is absorbed and the state changes to iM,m> is given
by
hv+ = g.BB + Airv 3 . 4
Here iTi " il/2 • This gives irise: co cwo resonance lines
separated by A from, each other, or by A/2 from B = hVQ/g!3.
Two satellite peaks in the ESR of CVD diamonds are indeed
observed as shown in Figure 3.7. The origin of the two
satellites will be identified and discussed in the next
section.
E. Electroliininescsiice Measurements
To investigate the electroluminescence (EL) from diamond
LEDs, a simple device was fabricated as shown in Figure 2.3.
114
M
+1/2
hv.
-1/2
/
hva.
J
hv_
m
-J./
-1/2
-1/2
•1/2
Fig. 3.6 Energy-level diagram for case S=l/2, 1=1/2. The
left side shows the Zeeman splitting only, the
right side shows the further hyperfine splitting.
The allowed ESR transitions are also indicated
115
< 1 1 i 1 1 1 1 1 i i i 1 i 1 1 : 1 • 1 ; i i j ' i —I—
Ns = 3.7E+17 gm~^
= 2.0026 -
CH.(0.5%) 4- H2(99.5%) ~
1 _
I 1 i 1 . 1 1 1 1 1 r 1 1 1 1
1 1
1 1
r 1
1 1
1 1
1
3.3 3.32 3.34 3.36 3.38 Magnetic Field (kG)
Fig. 3.7 ESR spectrum of diainond film (sample B) . The tv;o
satellites indicates hyperfine interaction between
dangling bond and nuclear spin
116
The EL spectrum and its relative intensity were measured using
an experimental set up as shown in Figure 3.8. The diamond LED
was driven by a DC power supply with voltage range 0-400
volts, and the emitted light from the LED passed through a
chopper and a monochromator (M) to a photomultiplier tube
(PMT). The signal was then sent to a lock-in amplifier, and
the spectrum was obtained by scanning the monochromator from
300 to 650 nm.
117
i
chopper
n PMT I 1
i m i ! I
• f
i
l -1
Lock-in amnjifif^r
Fig. 3.8 ExpejTiTusnta 1 set up for EL measuireinent
118
IV. RESULTS AND DISCUSSION
A. Electrolziminescence
EL was observed on both undoped and nitrogen doped
diamond films. However, the strongest emission was from an
undoped film deposited under regular conditions as described
earlier. Figure 4.1 shows a typical EL spectrum of an LED
fabricated from a 12 lam thick undoped film deposited over 24
hours- The broad blue-green emission band peaking at -450 nm.
has been observed by several groups and is called "band A"
31,32 « •
The LED characteristics varied widely among the devices.
The lifetimes varied from a few seconds to 10 minutes and
became longer as the electric field decreased. As the applied
voltage was increased, the luminance was enhanced and showed a
rapid increase above 2 00 volts. The dependence of the
luminance of a diode on the applied DC voltage is shown in
Figure 4.2. It is actually not clear whether the emission is
from the bulk of the diamond layer or originates from other
regions such as the interface between diamond and the metal or
the ITO. However, the peak position and the shape of the
spectra are very close to these of natural diamond-based LEDs
obtained by Prior and Champion.According to their
observations, blue light is emitted from the interior of the
119
1 i i 1 I i r 1 r
6 *—
4 —
2 —
l»
400 500 600
Wavelength (nm)
Fig. 4.1 Electroluminescence spectrum of diamond LED
120
iOO 200 300 400 500 T "J t Voltao'e o ~ \ • /
Fig. 4.2 The EL intensity vs applied DC voltage
characteristics of the LED
121
region. Nitrogen and some other defects in the films are
believed to serve as emission centers. Even though the films
were not intentionally doped, minute amounts of nitrogen were
probably incorporated into the films because the base vacuum
prior to deposition was higher than 10"^ torr.
B. The Native Defects of CVD Diamond Films
The nature of the defects in CVD diamond is very
important for the electronic and optical applications of these
films. In this work we identified for the first time and
reported^"^ a new defect center in CVD diamond manifest in the
hyperfine-split satellites in the ESR spectrum.
As mentioned earlier, we always observed two partially
resolved satellites in the ESR spectra, as shown for sample B
in Figure 3.7. The possible candidates responsible for the
hyperfine splitting are either (1.1% natural abundance) or
^H. They each have nuclear spin 1=1/2 and both participate in
the deposition processes. To identify which one is responsible
for the splitting, we first fit the ESR spectrum with a broad
Lorentzian, a narrow Lorentzian, and two off-center
Lorentzians as shown in Figure 4.3. The agreement is excellent
and indicates that the satellite intensity is -4% of the total
intensity. Since has 1.1% natural abundance and the
satellite intensity also varies from sample to sample, we can
rule out the oossibility of ^^c. However, we also prepared a
122
3.38 3.3 3.34 3.36 3.32 Magnetic Fieid (kG)
4.3 Fitting of the ESR spectrum by four Lorentzians
(solid lines). The dotted line is the sum of the
four Lorentzians
123
V = 9.53 GHz Ns = 3.6E-i-17 gm ^ g, = 2.0026 J ^^CK.(0.5%) ^ H2(99.5%) j
f*
3 o cj
_9
-4
3.25 3.3 3.35 3.4 Magnetic Field (kG)
Fig. 4.4 The ESR spectrum of 100% enriched diamond film.
Note the x scale is 200 G as coinpared to 100 G in
the other spectra
124
100% enriched sample (sample E) using 0.5% 12CH4 diluted
with 99.5% H2 as the reactant gases. The broad ESR spectrum
(Figure 4.4) is clearly due to the hyperfine splitting between
the dangling bonds and their neighboring nuclei.
To prove that H is responsible for the two satellites, we
replaced the H by deuterium which has much weaker nuclear spin
moment (1=1) than Indeed, the two satellites completely
disappeared from the ESR spectrum of sample C (Figure 4.5)
prepared by using 0.5% CD4 diluted with 99.5% D2 as the
reactant gases. The satellites also disappeared from sample D
prepared by using 0.5% CH4 diluted with 99.5% D2 (Figure 4.6).
Previous NMR studies^^"^"^ have indicated that most of
the hydrogen in thin diamond films resides on the crystallite
surfaces, and is removed after about 2 hours annealing at
850°C or above. However, the ESR line shape and the spin
density of sample B did not change at all even after annealing
for 1 hour from 740 to 1100°C at steps of 30°C (Figure 4.7).
Finally, the sample was annealed for 20 minutes in an Ar flow
by an oxyacetylene rorch at T > 12 00°C. The resulting ESR line
shape rem.ained essentially unchanged but the total intensity
dropped by factor of -4.5 (Figure 4.8). The satellites' strong
resistance to the high temperature annealing suggests that the
carbon dangling bond-H centers responsible for these
satellites are deeply trapped either on internal surfaces, or
125
Ns = l.lE-rl8 gm '
I,-. = 2.0028
c
o o
-5
3.38 3.36 3.32 o.o
Msignstic Fisld (kG)
Fig. 4.5 ESR spectrum of sample C. prepared from 0.5% CD4 +
99.5% D2- Note that the two satellites have
completely disappeared
126
-• 1 • • ' • 1 ' • • • 1 ' '
A
, , ^
"1 p = 9.53 GHz / Ns = 1.4E-fl8 gm~'
" = 2.0028
/ CH4(0.5%) 4- 02(99.5%)
— / -^Q'rr I , , • , I . , , , I , , , , ! , , . . I
3.3 3.32 3.34 3.36 3.38 Magnetic Field (kG)
Fig. 4.6 ESR spectru-i of sample D, prepared from 0.5% CH4 +
99.5% D2- Note that the two satellites have
completely disappeared
127
- 9.53 GHz Ns = 2.7E-rl7 gm
Annealed at llOO^^C x Ihr
-I 1
3.3 3.32 3.34 3.36 3.38 Magnetic Field (kC-)
4.7 ESR spectrum of sample Bl, which was annealed for 1
hour at steps of 30°C from 740 to 1100°C. Note that
the line shape did not change
128
• • I • ' • ' I ' ' • ' i ' • • • t ' ' '
Ns = 5.2E-rl6 gm"'
Annealed at 1500°Cx20mir
3.3 3.32 3.34 3.36 3.38 Magnetic Field (kG)
4.7 esr spectrum of sample b2, which was annealed for
20 minutes in an Ar flow by an oxyacetylene torch
at t > 1200°c. Note that the line shape was
essentially unchanged but the total intensity
drooped bv factor of -4.5
129
even in the bulk tetrahedral diamond network. Indeed such
centers may be expected since H is needed during the growth of
each layer to stabilize the surface, and hence some H atoms
may be "left behind" during this process.
All of the ESR spectra except the enriched one are
fitted into either two or four Lorentzains. The fitting
parameters and ESR results are summarized in Table 4.1. The
central narrow and broad Lorentzains, invariably observed in
all films, have been attributed to exchange-narrowed clusters
and isolated carbon dangling bonds at spinless nuclei,
respectively.^^ The two deuterated samples (C and D) showed
increasing central narrow Lorenztians may due to the
unresolved hyperfine split dangle bonds.
The hyperfine coupling constant of the H-dangling bond
complexes is equal to the field splitting between the two
satellites, i.e., about 14.5 G. Their intensity indicates a
density of about 0.03-0.2 ppm (Table 4.1), i.e., they are a
minute fraction of the total H content, which is typical 0.1-1
at. % in CVD diamond.
As mentioned earlier, the H centers could be located
either on internal surfaces or in the tetrahedral diamond
network. The SEM images showed that the average crystallite is
several m.icrons across at the top of the films. Yet if the
hydrogen
130
Table 4.1 The dangling bond density of each sample Ng, the
fraction of each Lorentzian component I, and its
derivative peak-to-peak linewidth PP. Sample B is a
regular diamond film, C and D are deuterated films,
and E is the enriched sample.
Sample
Ns
loiVi
I ( % ) PP(G)
Sample
Ns
loiVi broad narrow sat broad narrow sat
B 2.7 83 13 4 7.8 2.6 2.9
B1 2.7 78 15 7 7.5 2.6 3.0
B2 0.62 81 10 9 5.8 1.9 2.7
C 11 56 44 8.7 2.8
D 14 65 35 10.1 2.9
E 3.6
131
is essentially confined to the crystallite surface, the NMR
spin count indicates that the average crystallite is less than
O.Siim across.^® This discrepancy may be resolved by the fact
that the crystallite size rapidly increases during film
growth. It therefore does not allude to the existence of
internal surface within the observed crystallites. However,
such internal surface and/or microvoids are to be expected
within the kinetic model described by Anthony.In this
model, the hydrogenated surface is constantly etched and
repassivated by H atoms, but occasionally a methyl radical is
attached to an exposed carbon dangling bond, providing the
fundamental diamond network growth step. Such dynamical
processes may yield considerable internal defect structures.
Fanciulli and Moustakas'^'^ have indeed recently invoked
multivacancies and (111) stacking faults to explain the
properties of the ESR of their diamond films. However,
hydrogen that resides on the surfaces of such voids which are
connected to the crystallite surface will be released within a
few hours at -S50°C. In addition, a dangling bond on a largely
hydrogen-decorated surface should generally be adjacent to
more than one H atom. We therefore turn to the model of
dangling bond-H centers which are embedded in the tetrahedral
network.
In the simplest approach, substitution of a C by an H
stotp. slioiild 2705122. 1 stct. ens ijond.
132
and an interacting system of three other dangling bonds. The
ground state of this system should probably be an unpaired sp^
electron slightly overlapping the reconstructed paired
wavefunction of other two. However, this model should be
verified by a calculation of the defect wavefunction which
should yield an hyperfine splitting in agreement with the
observed value.
133
REFERENCES
1. J. E. Field, The Properties of Diamond. Academic Press,
London, 1979.
2. F. P. Bundy, H. T. Hail, H. M. Strong, and R. J. Wentorf,
Nature 176, 51 (1955).
3. P. W. Bridgman, Sci. Am. 193, 42 (1955).
4. W- G. Eversole, US Patents 3,030,187 and 3,030,188 (1962).
5. J. C. Angus, H. A. Will, and W. S. Stanko, J. Appl. Phys.
39, 2915 (1968).
6. J. C. Angus and N. C. Gardner, US Patent 3,661,526 (1969).
7. J. C. Angus, Sint. Almazy 3, 38 (1971).
8. S. Matsumoto, Y. Sato, M. Kamo, and N. Setaka, Jpn. J.
Appl. Phys. 21, 183 (1982).
9. S. Matsumoto, Y. Sato, M. Tsutsumi, and N. Setaka, J.
Mater. Sci. 17, 3106 (1982).
10. M. Kamo, Y. Sato, S. Matsumoto, and N. Setaka, J. Cryst.
Growth 62, 642 (1983).
11. Y. Matsui, S. Matsumoto, and N. Setaka, J. Mater. Sci.
Lett. 2, 532 (1983).
12. S. Matsumoto, ibid. 4, 600 (1985).
13. S. Matsumoto, M. Hino, and T. Kobayashi, Appl. Phys. Lett.
51, 737 (1987).
14. K. Suzuki et al., Appl. Phys. Lett. 52, 728 (1987).
P. snd, Msssis^r, 2— ^2.989)
134
16. Y. Hirose and M. Mitsuizumi, New Diamond 4, 34 (1988).
17. Y. Hirose, Proc. of 1st intl. Conf. on New Diamond Sci. and
Tech.. Tokyo, Japan, Oct. 24-26, 1988.
18. L. M. Hansen, W. A. Carrington, J. E. Butler, and K. A.
Snail, Mat. Lett. 7, 289 (1988).
19. Y. Tzeng et al., Appl. Phys. Lett. 56, 134 (1990) .
20. K. E. Spear, Earth Miner. Sci. 56, 53 (1987).
21. K. E. Spear, M. Frenklach, A. Badzian, T. Badzian, and R.
Messier, Ceram. Eng. Sci. Proc. 9, 1095 (1988).
22. J. J. Lander and J. Morrison, Surf. Sci. 4, 241 (1966).
23. M. Frenklach and K. E. Spear, J. Matt. Res. 3, 133 (1988).
24. B. V. Deryagin and D. V. Fedoseev, Growth of Diamond and
graphite from Gas Phase. Ch. 4 Izd. Nauka. Moscow, USSR,
1977.
25. B. V. Deryagin and D. V. Fedoseev, Sci. Am. 233, 102
(1975) .
26. N. Setaka, Proc. of 10th intl. conf. on CVD. Edited by G.
w. Cullen and J. Bloche. Electrochemical Society,
Pennington, NJ 1987.
27. N. Fujimori, T. Imai, and A. Doi, Vacuum 36, 99 (1986).
28. K. Kobashi et al., Phys. Rev. B 38, 4067 (1988).
29. B. D. Cullity, Elements of X-ray Diffraction, 2nd ed.
(Addison Wesley, Reading 1978) .
30. J. Loubser and J. V. Wyk, Rep. Prog. Phys. 44, 1201 (1978).
31. Y. Taniguchi et al., Jpn. J. Appl. Phys. 23, L1S48 (1989).
135
32. M. Kadono et al.. Mat. Res. Symp. Peoc. 162, 358 (1990).
33. J. R. Prior and F. C. Champion, Proc. Phys. See. 80, 849
(1962) .
3 4. H. Jia, J. Shinar, D. P. Lang, and M. Pruski, Phys. Rev. B
48, 17593 (1993).
35. J. M. Redwing, T. W. Root, and W. F. Banholzer, J. Appl.
Phys. (submitted).
3 6. K. M. McNamara, D. H. Levy, K. K. Gleason, and C. J.
Robinson, Appl. Phys. Lett, 60, 580 (1992).
37. S. Mitra and K. K. Gleason, Diamond and Related Materials
2, 126 (1993).
38. M. Pruski, D. P. Lang, S. J. Hwang, H. Jia, and J- Shinar,
Phys. Rev. B 49, 10635 (1994).
39. T. R. Anthony, Mat. Res. Soc. Pro. 162, 61 (1990).
40. M. Fanciulli and T. D. Moustakas, Phys. Rev. B 48, 14982
(1993) .
136
GENERAL CONCLUSIONS
Semiconductor materials of a-Si:H, CVD diamond films and
the related device have been fabricated and studied in this
work. The of a-Si:H increased by order of four after
annealing at -3 00°C, and the film showed weak light induced
degradation. The SAXS determined microvoid content was found
to be strongly correlated with the IR determined H content,
and high K content films showed columnar microstructure while
low H content films showed spherical or randomly oriented
voids. In the studies of H diffusion in a-Si:H, we found for
the first time the negative dispersion parameter a, which is
contradictory to the widely invoked "multiple trapping" model.
The CVD diamond based LEDs were successfully fabricated,
and a blue-green electroluminesence peaked at wavelength of
-450 nm was observed from such a device. A new H center which
may be deeply trapped either on internal surface or in the
bulk of diamond was identified for the first time through the
hyperfine features in the ESR signal. Further calculation
should be able to show the exact configuration of this new H
center.
137
ACKNOWLEDGMENTS
I would like to thank my research advisor Dr. Joseph
Shinar for his guidance, encouragement, and support throughout
this work.
My sincere appreciation goes to Dr. Ruth Shinar for her
collaborative work on SIMS, Drs. Don Williamson and Yen Chen
for their SAXS measurements, Dr. Scott Chumbley and Fran Laabs
for allowing me to use the SEM, and Drs. Davis Johnston and
Lance Miller for allowing me to use the x-ray diffractometer.
I am also very grateful for the help and counsel of my
graduate committee: Drs. Bing-Lin Young, Kai-Ming Ho, Alan
Goldman, Hsung-Cheng Hsieh, Michael Tringides, and Che Ting
Chan.
I also thank the physics department and Ames laboratory
for making the facilities available. Ames laboratory is
operated for the US Department of Energy by Iowa State
University under contract no. W-74 05-eng-8 2.
Finally, I would like to thank my wife Li Li to whom this
dissertation is dedicated. Her love, patience, and
understanding has made this research possible.
Recommended