structural, optical and photoelectronic properties of improved pecvd a-ge:h
TRANSCRIPT
Journal of Non-Crystalline Solids 137&138 (1991) 803-808 North-Holland
] O U R N A L OF
l g-CRNgI E Section 14. Amolphous germanium (a-Ge:H) and SiGe alloys
STRUCTURAL, OPTICAL AND PHOTOELECTRONIC PROPERTIES OF IMPROVED PECVD a-Ge:H
William PAUL
Division of Applied Sciences, Harvard University, Cambridge, Massachusetts 02138, USA
The preparation conditions by PECVD for a-Ge:H with photoelectronic properties approaching those of a-Si:H are described. Emphasis is placed on the identification by TEM, SEM, NMR and other techniques of structural defects on a scale of 1 nm and larger, and their elimination by adjustment of deposition parameters. Evidence concerning the valence and conduction band densities of states is reviewed, and preliminary information on the extension of this approach to improvement of a-Si~_=Ge~ :H alloys examined.
1. INTRODUCTION
It is generally agreed that the photoelectronic proper-
ties of a-Sil_=Gex:H are poorer than those of a-Si:H?
Given that (1) the alloys are usually prepared under
conditions which optimize a-Si:H, that (2) they risk the
complication of spatial fluctuations of the GelSi ratio,
and that (3) the fundame0tal processes of transport and
phototransport are not well understood even in a-Si:H,
various groups have undertaken the study of the pure
a-Ge:H end-point. Material with good photoelectronic
properties has now been reported. Despite some dif-
ferences in detail, the general conclusions concerning
preparation parameters and optimized photoelectronic
properties appear to agree. Our own results 2 are quite
close to those of the Siemens group?
Several of our papers ~ have described the conclu-
sion from many structural studies that there is a two-
phase network of high atomic density islands and lower
density connective tissue in this class of materials. Oth-
ers have focussed on the photoelectronic properties
and have attempted to relate them consistently to the
structure. In a recent review, ~ we suggested that the
optical properties were attributable to islands of differ-
ent structural quality, but that the transport and pho-
totransport were heavily influenced by the amount of
highly defective tissue connecting islands. Here we at-
tempt a general assessment of (1) our understanding
of the medium range structural order on a scale of 1 nm
and greater, (2) the modifications of the PECVD prepa-
ration parameters found efficacious in producing a
material of low dangling bond and low void content,
(3) the evidence for a band structure with reduced den-
sity of states in the band tail and energy gap, (4) the
relation of our band parameters to those of other inves-
tigators and also to extrapolations from the immense catalog of work on a-Sil_=Ge.~:H alloys, and (5) de-
sired improvements on which future work might usefully concentrate.
2. STRUCTURE
We seek to produce homogeneous a-Ge:H with a
negligible amount of low density connective tissue.
(1) On a scale of about 2 nm and larger, the 2-phase
island/tissue structure is often revealed by TEM Films
of a-Si:H and a-Ge:H with no TEM microstructure on this scale have both been made by us, but the depo-
sition conditions are quite different? ,2 Figure 1 illus-
trates (a) a-Si:H deposited at 250°C from pure Sill4
at a low deposition rate under conditions minimizing
substrate bombardment by ions, (b) a-Ge:H deposited
from (GeH4+H2) under similar conditions to (a). Het-
erogeneity is clearly visible, and (c) a-Ge:H deposited
from (GeH4+H2) at an electrode at a plasma-caused
self-bias of -150 V with respect to ground, which is
expected to promote ion bombardment. No structure
is visible. SEM pictures of cross-sections of thick films
deposited underthe same conditions as for films (a)-(c)
verify the likelihood of columnar structure when hetero- geneity is visible in TEM.
It is necessary here to discuss briefly the geometry
of ourtwo reactors? They have two electrodes differing
0022-3093/91/$03.50 © 1991 - Elsevier Science Publishers B.V. All rights reserved.
804 W. Paul / Structural, optical and photoelectronic properties of improved PECVD a -Ge:H
in area by x2. An r.f. power of 30 W produces a bias
potential at the smaller electrode (the powered cath-
ode, C) of -158 V and at the larger (the anode, A) of
-18 V. All of the films of a-Ge:H and a-Sil_xGex:H with
featureless TEM pictures have been deposited at the
smaller electrode of either reactor. We must recog-
nize that we are limited in our description of the precise
plasma conditions. Figure 2 shows TEM micrographs
of a-Ge:H films deposited at the cathode for spacing
of the electrodes increasing from 1.0 to 3.2 cm 4. Al-
though the biases on both electrodes change by only a
few percent, the visible film heterostructure increases
markedly, and the photoelectronic properties deterio-
rate. Although the plasma conditions near the cathode
must have changed, the changes in the film are clearly
not caused by a change in ion bombardment energy.
This result differs from the conclusions of the Siemens
group for their reactor.
The structure observable in TEM is verified in sev-
eral kinds of less direct structural studies. An example
is that of typical infrared absorption spectra for cathode
(C) and anode (A) films. In addition to the monohy-
dride stretch mode at 1875 cm -1 and the wag mode at
560 cm -1 seen in environmentally-stable C films, the
A films have a stretch mode at 1975 cm -1, bending
modes at 760 and 830 cm -1, and show contamina-
tion by C and O which increases with exposure to the
atmosphere. Evolution of 14 on heating a sample at
a controlled rate may be used to estimate the total
H-content and also to provide information on how the
H is bound. In a-Si:H there are typically two peaks:
a low temperature one, independent of sample thick-
ness, indicates that there are H2 molecules present
which can leave the film easily, while a peak whose
(higher) temperature position depends on thickness is
a signature of atomic H from Si-H bonds which exit
by diffusion, with occasional trapping at Si dangling
bonds. The results for a-Ge:H are more complex. Our
C material also shows two major peaks, but the H2
evolving at low temperature can exist in different sites
only revealed by deuteron magnetic resonance (DMR)
(see below). The A material may have two additional
peaks: a very low temperature one verified by an RGA
to correspond to the evolution of contaminant gases
which enter the open-network structure after deposi-
tion, and a very high temperature one corresponding
to the evolution of GeOx and H2. Differential scanning
calorimetry shows that the C material, amorphous by
electron diffraction as-deposited, crystallizes close to
430°C in a discontinuous transition s reminiscent of ex-
plosive crystallization. The A material crystallizes in 1
or 2 stages depending on the temperature of deposi-
tion and the substrate used. 5 Sequential crystallization
at temperatures which may differ by 100°C, with a to-
tal heat of crystallization closely matching that found in
single event crystallization, is a clear (and new) signal
of a 2-phase network. The microcrystals formed after
the first incident of crystallization are of the order of
8 nm diameter; their properties remain to be explored?
(2) We now consider structure on a scale smaller
than about 2 nm. Very detailed information on the
environment of H atoms may be obtained from anal-
ysis of deuteron magnetic resonance spectra (DMR),
studied by Professor Richard Norberg on our a-Si:D,H,
a-Sil_xGe~:D,H, and C and A a-Ge:D,H produced by
substituting D2 for H2 in the preparation plasma. ~ At
least six environments for D (or H) are identifiable:
(1) tightly-bound Ge-D, (2) D2 or HD molecules in high
density, as in solid D~. These may occupy the tis-
sue regions visible in TEM, but also regions smaller
in size. (3) isolated D2 or HD molecules, probably
in very small voids or even interstitial positions in the
fully-coordinated CRN. (4) germanyl rotors, GeDxH3_x.
(5) weakly-bound D, giving a DMR signal identifiably
different from molecular species and from tig htly-bound
Ge-D and (6) para-D2 molecules. Comparison of the
line-shapes of the DMR spectra for C and A a-Ge:D,H
verify that there is more disorder in the environment of
Ge-D bonds than Si-D, and more disorder in the Ge-D
bonds for the A than for the C Ge. Several new and
significant conclusions emerge. One is that D2 and HD
exist in the C material both in the dense and isolated
forms. Thus this material must have void regions, invis-
ible in TEM, which are of larger dimension than those
in our state-of-the-art a-Si:D,H. This is consistent with
the observation of much low temperature gas evolution
W. Paul / SO~ctural, optical and photoelectronic properties of improved PECVD a-Ge:H 805
(a) (b) (c)
FIGURE 1 TEM micrographs for (a) a-Si:H (b) A a-Ge:H and (c) C a-Ge:H. See text for details.
in our C films and the conclusion from a comparison of
gas evolution, forward recoil scattering and infrared ab-
sorption measurements that the total H-concentration
is about twice the bonded component. Another impor-
tant conclusion is that of the existence of weakly-bound
single D at the 10 -4 concentration level. While to date
this has been investigated mostly in a-Si:D,H films, its
presence in a-Ge:D,H also is considered highly likely.
We conclude from the DMR studies that even the C a-
Ge:H has void structure on a scale below that seen in
TEM or appreciated in GE, IR and DSC investigations.
This void structure has also been verified by SAXS,
which indicates that the C material has many fewer
voids than the A, but that the void distribution in it is
much more extensive than in state-of-the-art a-Si:H. 7
Thus it is evident that structural improvements on the
scale below 2 nm are still necessary.
3. COMMENTS ON PECVD PARAMETERS
These results establish that, in our reactor, differ-
ent preparation conditions are required for "good" a-
Ge:H than for a-Si:H. This agrees with the conclusions
of other groups which use both PECVD and sputter-
ing. However, it appears that there may be different
combinations of parameters which all produce reason-
able material. Matsuda and Tanaka very early recom-
mended heavy H2 dilution of the plasma. 8 We have
found that attempts to produce a-Ge:H from undiluted
GeH4 result in dust rather than a coherent film, and that
dilution with H2 gives films of good quality. However,
we have also produced good quality C material from
(GeH4+He) plasmas, which suggests it may be the di-
(a) (b) (c)
FIGURE 2 TEM micrographs for C a-Ge:H deposited at several electrode spacings D (a) D = 1.0 cm (b) D = 2.0 cm (c) D = 3.2 cm.
lution rather than the specific addition of H2 which is
responsible for the improvement.
Karg et aL 3 have correlated photoelectronic qual-
ity with an ion bombardment parameter. We too have
found that increasing r.f. power leads to increased sub-
strate bias at the cathode and also to improved pho-
toelectronic properties over the range of power used
by Siemens. However, the improvement in proper-
ties was found to pass through an extremum and to
decrease as the power increase was continued. More-
over, we have found improvements (see Figure 2) in
the structure and photoelectronic quality of our films
when the substrate self-bias remained unchanged. We
conclude that the geometry of the apparatus and the
distribution of fields in it are important, as well as the
gas and power parameters. Until studies of the plasma
and the substrate-piasma reactions are completed, it
appears that empirical optimization of individual appa-
ratuses will be needed. Nevertheless, the departures
from Sill4 conditions are a guide: while for a-Si:H it is
preferred to keep ion bombardment to a minimum, for
a-Ge:H it appears to be better to arrange the geometry
and increase the power so as to promote bombard-
ment, and to add H2 or He. (The function here of H.2, or
HF, may be to etch weakly-adhering material, but this is
not proven). Recently, Doyle etal . have confirmed that
there are major differences in the radical interactions
in GeH4 and Sill4 discharges, and they have proposed
tentative reasons why these differences lead to poorer quality a-Ge:H films. 9
806 W. Paul~ Structural, optical and photoelectronic properties of improved PECVD a-Ge:H
TABLE 1 Summary of properties of a-Ge:H films deposited at the cathode and anode at 150°C. The illumination used to determine zXI was 8 x 10 is photons/cm2sec at 1.25 eV.
Sample Eo4 (eV) E03 (eV) r~(2.0 tim) r/fiT- (cm2/V) zXI/Id E,, (eV) ~o (~ cm)
Cathode 1.25 1.10 4.03 3.0x10 -7 1.1 x l 0 -1 0.62 1.0x10 s
Anode 1.24 1.07 3.78 4.2x 10 -1° 1 .0x l0 -~ 0.52 3.1 xl03
Sample Eo,Pos o~(0.7 eV) EpL z~EpL Ns C,v,zR stress (meV) (cm -1) (eV) (eV) (cm -3) (at. %) (kbar)
Cathode 51 8.3 0.81 0.19 5x1016 6.4 +4.0
Anode 89 91 unmeasurable unmeasurable 6 x 1017 5.6 -1.5
It is plausible to extrapolate that the best quality
a-Sil_,Gex:H alloys cannot be prepared from a sin-
gle plasma, and that gentle deposition techniques (re-
mote plasma CVD, photo-CVD, hot-wire) may not be
as appropriate for Ge as for Si. These inferences may
change when we understand better the roles of plasma
interactions and substrate-plasma processes. Control
of the surface mobility of depositing radicals is clearly
important.
4. PHOTOELECTRONIC PROPERTIES
The photoelectronic properties, reported at greater
length elsewhere, 2 consistently confirm that mate-
rial with less heterostructure has sharper absorption
edges, smaller Urbach tail coefficients, less subband-
gap absorption, better photoluminescence and supe-
rior qffT products. Table 1 lists an earlier comparison
of C and A films which we shall use as illustration here,
with some addition of statistics: (1) For typical C and
A films deposited under identical conditions, the E04's
are about the same, implying that the joint valence-
conduction band density-of-state (DOS) distributions
are the same at this energy separation. (2) The differ-
ence (Eo4 - Eo3) is about constant at 0.15 eV for all
C films; this is, however, smaller than for the A films,
and it is also slightly smaller than we found in our ear-
lier study of a-Sij_~.Gex:H alloys. See Figure 3. The
fact that the upper part of the absorption edge is only a
little different in C and A materials with distinctly differ-
ent heterostructures implies that the islands dominate
this part of the absorption edge. (3) The Urbach pa-
rameter E0 is much smaller in the C material; for 16
films the PDS-determined values lie between 42 and
61 meV and the CPM-determined ones range from 40
to 55 with 15 of the 16 below 50 meV. Thus the va-
lence band tails are as sharp as in a-Si:H. This fits
earlier reports that Eo is roughly constant through the
a-Si~_,~Gex:H alloy series; we have recently also con-
firmed this conclusion for our a-Sij_,:Ge,::H alloys at
the Ge-rich end, deposited under conditions similar to
those used for a-Ge:H. The larger values of £o in the
A films are attributed primarily to a deterioration of the
island material and secondarily to the poorer tissue.
The subband-gap absorption coefficients at 0.7 eV for
C films are about 10 cm -1 by PDS and x 2-3 smaller
by CPM measurements. This is consistent with the
poorer a-Ge:H structure below 2 nm and is a factor 10
larger than in good a-Si:H. (5) The transport parame-
ters E,, and ,7o are not very informative regarding the
differences between C and A material, yet there is a
signature, not yet understood, which consistently dis-
tinguishes the two: the lno versus ;/T relation for A
material is linear down to 25°C while that for C mate-
rial is always concave to the 1/T axis, with no sign of
a sharp kink at any temperature. Moreover, the o-0 is
always larger in C films for the same E~; we interpret
this to mean that the shift with temperature of the Fermi
level in C films is much greater, probably because it
is sited in a low and rapidly varying part of the gap
DOS versus energy relation. (6) The A films have no
photoluminescence we could measure, but the C ma-
terial (38 films) have E(peak) = 0.80±0.02 eV; FWHM
~E = 0.19±0.02 eV and intensity 10 2 to 10 -3 of
a-Si:H. The PL is almost independent of variation of
W. Paul~Structural, optical and photoelectronic properties of improved PECVD a-Ge:H 807
TABLE 2 Correlation of q/.~T with subband-gap absorption coefficient for A a-Si:H, C and A a-Ge:H and a-SiGe:H.
Eo4 (eV) E(eV) e(cm -~ ) q#T(cm~/V) c~q#T(cm/V)
a-Si:H 1.9
C a-Ge:H 1.2
A a-Ge:H 1.2
a-SiGe:H(1988) 1.4
1.2 0.5 5x10 -6 2.5×10 -6
0.7 14(16 films) 2.5x 10-~(16 films) 3.5x10 -6
0.7 81(6 films) 9× 10-1°(6 films) 7.3x10 -s
0.8 20 5x10 -1° l x 1 0 -s
deposition parameters, and is different from extrapola-
tions from our and Carius 1° earlier measurements on
a-Sil_=Gex:H alloys. See Figure 3. (7) The spin den-
sity in C material is x 10 smaller than in A material, but
x ]0 larger than in a-Si:H. This is consistent with the
differences in subband-gap absorption. (8) Finally, the
most dramatic element in Table 1 is the superior 71/~- of C films.
A second illustration involves the photoresponse of
C films from reactor 2 at the several electrode spacings
used for the TEM's of Figure 2. Figure 4 shows that
zM/I (dark) product decreases monotonically as the
electrode spacing is increased, which we link directly to
the increased heterostructure in Figure 2 (the reasons
for these changes are not relevant here).
In both of our reactors the photoelectronic properties
change as the deposition parameters are varied. Al-
though this delineates the "best" volume of parameter
space for our reactors, it is not directly transferable to
others. What is more significant are the correlations of
the properties among themselves. Thus, an improved
microstructure evidenced by TEM (C films) correlates
without exception with decreases in Eo and ~ (0.7 eV)
and increases in q#T. Three DMR parameters - - the
fraction of tightly-bound D (i.e., H), the fraction of tightly-
bound D2 or HD, and the inverse relaxation rate, which
is proportional to the density of paramagnetic defects,
also correlate extremely well with the 7/#T found for identically-deposited films. 6
It is interesting to attempt a correlation between q#T
and the subband-gap ct., which reasonably represents
the subband-gap DOS determining T. The results are shown in Table 2. From the equivalence of a.q#T for a-
Si:H and C a-Ge:H we infer that the reduction in ~//~7- in
2.0
1.5
> v
>, 1.0
E LU
0.5
, . . . . . . . . . . ,
- • • O
E0~ A Z~ ~
• []
m r q %
Epl"
AEpl
• • O@o ¢)d9o
00. , , , I , , , l l l l l l l l l l l r
0.0 0.2 0.4 0.6 0.8 X
.0
FIGURE 3 Dependence on Ge content of Eo.~, E,~3, £p~ and ,AEp~. Solid symbols, 1988; open symbols 1991
0.05
0.04
,~ 0.03
0.02
0.01
0 1.0
~ , l l , ~ , , i , ~ , ~ l ~ , , , I
Q
• |
r , , , l , , , r l . . . . I . . . . I • t
1.5 2.0 2.5 3.0
D (cm)
3.5
FIGURE 4 AI/I (dark) versus electrode spacing D.
808 W. Paul~Structural, optical and photoelectronic properties of improved PECVD a-Ge:H
the latter is directly related to the increase in gap DOS
represented by c~ (0.7 eV). The much smaller c~71ffT
for A a-Ge:H we link to the occurrence of more mi-
crostructure (tissue). In the fourth row we have listed
data on a-Sil_=Ge=:H from our earlier studies; here
too, the reduction in qff-r is greater than expected from
(0.8 eV); and again we note that TEM clearly shows
the existence of a heterostructure. It is also interesting
to attempt a correlation between our present results
and extrapolations to pure a-Ge:H of data on SiGe al-
loys. Aljishi et aL ~ have reported the variation of their
subband-gap DOS with optical gap for a series of a-
Sil_xGex:H,F alloys. Extrapolation to an optical gap
of 1.1 eV (the value for our cathode Ge) suggests a
gap DOS of ~ 101~ cm -3 eV -~. This is more than an
order of magnitude larger than our estimates for our
material. Aljishi et al. suggest appropriate parameters
for both the entropy model and the kinetic model for
defect equilibration which fit their alloy data. Obviously
these suggested parameters do not fit our results for
pure a-Ge:H.
5. BAND STRUCTURE OF a-Ge:H
The similarity of the valence band and valence band
tail structures (Eo4, £03, Eo) in C a-Ge:H and a-Si:H is
very clear. By contrast, the conduction band tail struc-
ture seems to be more complicated, since our time-of-
flight data taken at 25°C invariably show anomalous log
current-log time characteristics. Our EPR experiments
suggest an increase in the density of dangling bonds
by x l0 over a-Si:H and, from our crude analysis of
~q#~-, this increase in gap DOS is sufficient to explain
the decrease in qff.r. Our structural studies suggest
that point and extended defects on a 1 nm scale could
be responsible for the increased gap DOS. However,
we note also the suggestion by Shu Jin and Ley v-' of an
H-related, non-DB defect, possibly a Ge-H-Ge 3-center
bond complex. Although an intriguing speculation, one
is reminded that the energies of the antibonding states
of much simpler Ge-H bonds are assumed with little
justification to lie in the conduction band.
6. CONCLUSIONS
(1) Different PECVD conditions are required to pre-
pare a-Si:H and a-Ge:H of comparable photoelectronic
quality.
(2) Ion bombardment and/or etching appear to be
necessary to eliminate heterostructure in a-Ge:H.
(3) Structural (and possibly chemical) inhomogene-
ity on a scale of 1 nm still exists, whose elimination
should reduce the gap DOS.
(4) Our best a-Ge:H has an absorption edge as
sharp as that of a-Si:H, but a gap DOS still x 10 larger,
which is adequate to explain the lower qy~-.
(5) Future work should seek to understand better
the DOS in the CB tail, and to extrapolate the methods
used here to Ge-rich a-Sil_,,Ge~.:H.
ACKNOWLEDGMENTS
This work was financially supported by the SERI un-
der Contract XX-8-18131-1. I thank Professor J.H.
Chen, Dr. W.A. Turner, Dr. F.C. Marques, Mr. B. Bate-
man, Mr. S.J. Jones, Ms. D. Pang, Ms. A.E. Wetzel
and Mr. P. Wickboldt for their whole-hearted collabora-
tion in our research.
REFERENCES
1. An extensive set of references is given in W. Paul et al., Proc. Mat. Res. Soc. 219 (1991) 211, to be published.
2. Reference 1. Also W.A. Turner et ai., J. Appl. Phys. 67 (1990) 7430.
3. RH. Karg et al., J. Non-Cryst. Solids 114 (1989) 477; also, to be published in Solar Energy Materials (1991).
4. P. Wickboldt et al., Phil. Mag., to be published. 5. W. Paul et al., Phil. Mag. B63 (1991)247. 6. Reference 2. Also R.E. Norberg et al., this volume. 7. R. Crandall, private communication. 8. A. Matsuda and K. Tanaka, J. Non-Cryst. Solids
97/98 (1987) 1367. 9. J.R. Doyle et al., J. Appl. Phys., 69 (1991) 4169.
10. R. Carius in Amorphous Silicon and Related Mate- rials, ed. H. Fritzsche (World Scientific, Singapore 1989) p. 939.
11. S. Aljishi et al., in Amorphous Silicon and Related Materials, ed. H. Fritzsche (World Scientific, Singa- pore 1989) p. 887.
12. Shu Jin and L. Ley, Phys. Rev. B44 (1991) 1066.