Structural study of bismuth films and its consequences
on
their electrical properties
J.
Buxo, M.
Saleh and G.Sarrabayrouse
Laboratoire d’Automatique et d’Analyse des Systèmes du C.N.R.S., 7, av. du Colonel-Roche, 31400 Toulouse, France
G. Dorville
Société Jules-Richard, 116, quai de Bezons, 95 Argenteuil, France
J.
Berty
and M. BrieuLaboratoire de Physique Structurale, Equipe de Recherche Associée au C.N.R.S., Cinétique Cristallochimique des Couches Minces, Université Paul-Sabatier,
118, route de Narbonne, 31077 Toulouse Cedex, France
(Reçu le 13 mars 1979, révisé le 4 février 1980, accepté le 8 février 1980)
Résumé. 2014 Les auteurs étudient les
propriétés
structurales des couches minces de bismuth à 1’aide destechniques
suivantes : diffraction des électrons par réftexion, diffraction des rayons X,
microscopie électronique
parbalayage
et diffraction et
microscopie électronique
par transmission. Ilsrappellent
lespropriétés électriques
des couches minces de bismuth (mesures de conductivité et du facteurRH). Il
apparait que les états de surfacejouent
un rôledéterminant dans les définitions des
propriétés
du facteur RH et de la conductivité. Cette conclusion est étayéepar les résultats de 1’etude structurale. Une
propriété importante
du mécanisme de conduction telle que le libre parcours moyen des porteurs a été estimée et sa valeur est aussi en accord avec les résultats de l’étude structurale.La densité des états de surface ainsi que la hauteur de barrière associée aux joints de
grain
ont été évaluées enfonction des différentes techniques de
preparation.
Abstract. 2014 The structure of bismuth thin films prepared according to three different
deposition techniques
isinspected
by : reflection electron diffraction, X-ray diffraction, scanning electronmicroscopy,
transmissionmicroscopy
and electron diffraction. The electrical conductionproperties,
RH and conductance measurements, of these films are then discussed. It appears that surface statesplay a major
role indictating
theproperties
of both RH andconductivity.
This is clearly confirmed with the aid of the conclusions of the structuralstudy.
Bycomparing
size effect theories with the
conductivity
data the carrier mean free path can be estimated to be of the order of 1 000 Å in good agreement with the observation of the film structure. The surface statedensity
and the energy barrier that controls the carrier transport across the grain boundaries have also been evaluated as a function of the preparationtechnique.
Classification
Physics Abstracts
73 . 60D
1. Introduction. - Some electrical and electro- mechanical
properties
of thin bismuth films werepresented previously [1].
Theannealing process
andthe substrate
temperature during evaporation
haveappeared
to have an influence on theseproperties.
Inorder to get a
deeper insight
into thephysical
pro-perties goveming
thesephenomena,
asystematic
structural
study
of the filmscorresponding
to threedifferent
processing
methods has been carried out.The conclusion is that the
annealing
process(pro-
cess n°
2)
does notbring
about anysignificant
struc-tural
modification ;
the films arehighly
textured withcrystallite
size dimensionlying
between1000 Å
and2 000
A. Nevertheless,
films grown on hot substrates show a markedtendency
to coalescencethereby giving
rise topolycrystalline grains
of verylarge
dimension. The same conclusion is reached when two different substrates are
used, namely,
a resinelayer (for
thestudy
of thickfilms)
andamorphous
carbon(for
thinnerlayers).
The
preceding
conclusions have madepossible
toobtain some new and
quantitative
features of theArticle published online by EDP Sciences and available at http://dx.doi.org/10.1051/rphysap:01980001505096100
electrical
properties
of bismuthfilms, particular
attention is
paid
to the Hall Constant(RH)
and con-ductivity (0)
measurements.The effect of thermal treatments either after
depo-
sition
(process
n°2)
orduring deposition (process
n°3)
on the
dependence
ofRH
ontemperature
is discussed in detail. The structuralstudy
has shown that nomarked structural différence is obtained after annea-
ling
which puts forward themajor
role of surface states to accountfor,
the influence of theannealing
treatment on the
conductivity values,
on the onehand,
and the marked différences between theRH(T)
curves for the three différent processes, on the other hand.
The
conductivity
variation with the film thickness has beencompared
with both the surfacescattering theory (Fuchs-Sondheimer) [21] and
thegrain
boun-dary scattering theory (Mayadas-Shatzkes) [22].
Ithas been found that the
conductivity
of films pro- cessedaccording
to process n° 2 can be accounted forby
thepredictions
of the F-Stheory
with themean free
path
of the order of 1000A
which is ingood
agreement with thegrain
size estimate from the structuralinvestigation
with the aid of theScanning
Electron
Microscope. Nevertheless,
films from pro-cesses n° 1
and
n° 3 cannot bereasonably
describedby
thepreceding
theories and this appears to be due to the presence ahigh density
of surface states.Indeed,
thecomparison
between theconductivity
values of films from
processes
n° 1 and n° 2 has allowedto evaluate this surface states
density
which for films of process n° 1 appears to be of the order of 2 x1012 cm-2.
With an associated energy barrierheight
of 1.3 eV. Theannealing
treatmentbrings
about a reduction in the surface states
density
whichin turn causes the
barrier energy
values to decrease.The positive
value of thetemperature
coefficient of theconductivity - 1 (semiconductor-like) for
films from process n° 1 can then be
interpreted
interms of the
increasing probability
ofovercoming these
barriers at thegrain
boundaries.Conversely
the observationaccording
to which the sizeof
thedisoriented
polycrystalline grains
increaseswith
thick-ness for films from
process
n° 3 is consistent withthe experimentally
observedtendency
for themobility
to decrease with thickness
[1].
2.
Experimental
devices. - The structuralstudy
has been carried out on
squared
aluminium substratesof dimensions 40 x 40 mm. An
insulating
10 ym thick resinlayer
wasdeposited
on thesubstrate using
the method of
auto-regulated electrophoresis
pre-viously
described[1].
Bismuth films of thicknessescomprised
between 1 000A
and 3 000A
weredepo-
sited onto the resin
layer by’
thermalevaporation
under vacuum of
10-6
torr, at a rate ofevaporation
of 50
Â/s.
Thepurity
of the metal used is 99.999%.
Different thermal treatments were
given [1]
andtheir effects
with regards
to their structuralproperties
have been
carefully analysed by using
thefollowing
three
techniques :
1)
Reflection electron diffraction.2) X-ray
diffraction.3) Scanning
electronmicroscopy.
The
experimental
conditions ofevaporation
werethose
already
described in[1]
that we recall hereafter :A fourth
technique
is used for the structuralstudy
of bismuth films of
small
thicknesses(100-400 A)
prepared by
thermalevaporation
carried out in theinterior of an electronic
diffractograph
asdepicted
infigure
1. The substrate materialis,
in this case, amor-phous carbon,
and the rate ofevaporation
50Â/s.
Fig. 1. - Diffracto graph object chamber bolt-equiped.
By using
a heatedsubstrate-holder,
it hasbeen
pos- sible to obtain the twofollowing preparation
tech-niques :
1)
Condensation on a substratekept
at room tem- perature.2)
Condensation on a substrate heated at 100 °C.These
experimental
conditions are similar to those of the aforementioned films ofgreater
thicknesses.3. Structural
study.
- This sectiondescribes the
experimental
resultsobtained by using
thedifférent
techniques
ofanalysis
as well as the conclusions obtained withregards
to the films structure.3.1 REFLECTION ELECTRON DIFFRACTION. - The films have been examined inside an electronic dif-
fractograph operating
at an accelerationvoltage
of80 kV.
Figure
2a shows the diffraction pattern cor-responding
to process n° 1. Thediagram
has beencompletely
indexed as shown onfigure
2b. Thisindexing
as discussed inAppendix, clearly
shows thateach
point
in the bismuthreciprocal
lattice hasactually
been found onfigure
2a. Thequality
of thepicture
allows to conclude that the bismuth film is made up ofmicrocrystallites
that arrange themselves in ahigh quality
texture. Identical results wereobtained for
films prepared according
toprocess
n° 2.Fig. 2. - a) Reflection electron diffraction pattem for Bi films,
3 000 A thick. Preparation process n° 1. b) Indexed diagram of (a).
It should be noticed that the
diffractioii-diagram
for films
prepared according
to process n° 3 and shownon
figure
3 presents a different aspect, the presence ofrings
should beinterpreted
asbeing
causedby
thegrains
withcompletely
random orientation.3.2 X-RAY DIFFRACTION. - This
study
was rea-lized with the aid of a CGR
type
10 diffractometeroperated
at 40 kV and 10mA.
Therecordings
of theFig. 3. - Reflection electron diffraction pattem for Bi films,
3 000 A thick. Preparation process no 3.
curves of azimuthal distribution were carried out
by using
theKai
radiation of copper whosewavelength
is  = 1.540
5 A.
Fig. 4. - Intensity of diffracted X-ray for 3 000 A thick Bi films.
a) Preparation process n° 1. b) Preparation process n° 2. c) Pre- paration process no 3.
The
sample
holder turns in a verticalplane
with aspeed
of 1degree/min.
The diffracted intensities weregraphically
recorded and are shown onfigure
4 forthe three different processes. The Y-axis shows the diffracted
intensity
and the X-axis the azimuthalangle
indegrees.
Table I summarizes the obtained results.Table 1.
It should be noticed that the films
corresponding
to processes n° 1 and n° 2
only display
the0003, 0006, 0009, ...
reflections which aretypical
of aperfect
texture, whereas for process no 3 a
supplementary
reflection
corresponding
to the1012
direction is observed in addition to thepreceding
ones. Table IIsummarizes :
a)
The different hkl reflections of bismuth as des- cribed within ’the threecrystallographic
systems,namely, hexagonal,
rhombohedric andpseudo
f.c.c.b)
The values of theinterplanar spacing (dhkl
calc. > 1.118À)
and those of the different intensities as calculated within the kinematicapproxi-
mation for electron diffraction
(the
value 100 has been attributed to thelargest reflection).
c)
The values 0 of theBragg angle
as calculatedfor
X-rays by using
theKa
radiation ofCopper.
The
comparison
of the calculated values in table II with those foundexperimentally
andpresented
intable 1 is very
satisfactory.
The
X-ray study
shows therefore that films from process n° 1 and 2display
aperfect
texture, whereas those from process n° 3 areprobably
made up ofcrystals
of random orientationssuperimposed
oncrystallites showing
the aforementioned texture.Table II.
3 . 3 SCANNING ELECTRON MICROSCOPY. - The bis- muth films were also examined with the aid of JEOL JSM-25 S
Scanning Microscope.
Eventhough
the resolution related to this
technique
isfairly
weakit is nevertheless
possible
to obtain information about theshape
and the dimensions of thegrains
orcrystallites
that make up the films.Figure
5 showsthe results for films 1 500
À
thick.Fig. 5. - Scanning electron micrograph for 1500 A thick Bi films.
a) Preparation process n° 1. b) Preparation process n° 2. c) Pre- paration process n° 3.
For films
corresponding
to process n° 1 and n° 2grains
ofhomogeneous shapes
withgrain
size of theorder of 2 000
A (Fig.
5a andb)
were observed. Forfilms
corresponding
to process n° 3(Fig. 5c)
twotypes
ofgrains
can be observed :-
homogeneous grains
ofshape
identical to thosein
figure
5aand b,
-
larger grains
with thicknesses aslarge
as 6 000Â.
3.4 INVESTIGATION OF BISMUTH FILMS OF SMALL THICKNESSES
(100-400 Å)
BY TRANSMISSION MICROS- COPY AND ELECTRON DIFFRACTION. -Figure
6 showsthe
micrographs
and thecorresponding
diffractiondiagram
for the twopreparation techniques
describedin section 2 in connection with their
particular
inves-tigation procedure.
Fig. 6. - Investigation of Bi films of thickness 100-400 À by transmission microscopy and electron diffraction. Technique 1 : a) microphotography, b) diffraction pattern. Technique 2 : c) microphotography, d) diffraction pattern.
For films condensed on a substrate
kept
at room temperature themicrograph
infigure
6a shows thatthe film is continuous and is made up of
grains
ofpolygonal
forms with dimensionscomprised
between1 000 and 2 000
Á.
Thecorresponding
diffractiondiagram
shown onfigure
6b indicates that the filmsalmost
exclusively
consist ofgrains presenting
thetexture
(0001).
For films condensed on a hot
substrate,
the micro-graph
shown onfigure
6c shows the presence ofgrains
identical to those in the
micrograph
onfigure
6atogether
with otherlarge crystals
of diameter about 4 000Á.
The appearance of these lattercrystals
canprobably
be due to thehigh temperature
of the substrate which causesduring
the condensation process, the coalescence of certain zones of the films.We should also notice on the same
micrograph
thepresence of white zones
surrounding
everylarge crystal
which can beinterpreted
as due to somediscontinuities in the films as a result of the coalescence process.
All
rings
onfigure
6dcorrespond
to thecompletely
disoriented
polycrystal.
If thecomparison
is madebetween
the calculated and theexperimental
intensi-ties,
it appears that the reflectionscorresponding
tothe texture i.e.
(1120, 0330, 2240,...) give
rise tolarger experimental
intensities. It should then be concluded that whendepositing
bismuth on a hotsubstrate,
two
types
ofcrystallites
are obtained : a first lot ofgrains
withplanes (0001) parallel
to thesubstrate
andanother lot of
completely
disorientedcrystals of larger
size.
In conclusion the three
types
of filmpreparation
allow the
following description
of the film structuralproperties :
1. Process n° 1 leads to films
which always present
a
high quality
texture withhomogeneous grains
ofpolygonal
form. For thicknesses of the order of 1 500A
thegrain
size isapproximately
2 000Á.
However it appears that the
grain
size tends toincrease with the film thickness.
2. Process nD 2
gives
results identical to those obtainedfor process
n° 1.3. Process n) 3 leads to two
types
ofcrystals : a) homogeneous grains
withperfect
texture iden-tical to those in process n° 1 and n°
2,
b) grains
oflarger
dimensionsreaching
10 000Á
for the
thickest
films. Thesegrains
arerandomly
oriented with
respect
to each other.These
conclusions are also confirmedby
the iden-tical
results obtained from thecomplementary
struc-tural
study
realizedby
transmissionmicroscopy
andelectron diffraction on films of lower thicknesses.
, These
results are in agreement with thosereported by
numerous authors[2-11]. Indeed,
it hasappeared
that the
(0001) planes
tend to lieparallely
to thesubstrate whenever the films are
evaporated
on anamorphous
substrate ofcarbon,
collodion orglass
and also when
they
areepitaxially
grown onboth,
mica and sodium chloride.Contrarily,
ifepitaxially
grown on
potassium
chloride orpotassium
bromidesubstrates the
deposit
ofplanes (0112) parallel
tothe substrate seems to be enhanced.
4.
Conséquences
of the structuralstudy
on someconduction
properties
of bismuth films. - In whatfollows,
ageneral
discussion of someimportant
electrical characteristics of bismuth films obtained
by
different authors is
presented.
Theinterpretations
willbe
compared
to the results of thepreceding
structuralinvestigation.
This discussion leads to a new conduc- tion model and to a betterunderstanding of
theannealing physical
mechanism.4. 1 CONDUCTIVITY. THE « SURFACE STATES » MODEL.
- The main
properties
of theconductivity
of thinbismuth
films
can be summarized as follows :i)
Its value isproportional
to the film thickness[l,14,16-19].
ii)
Theconductivity
valueincreases
for films inclucl-ing
thermal treatments(process
n° 2 and no3) [1, 16].
iii)
Thetemperature
coefficientof
theconductivity
is
positive (semiconductor-like)
up until acritical temperature
valueTc
where itchanges
intonegative
values
(metal-like) passing through
zero[1, 14, 16, 20].
The value of
Te
decreases for films of process n° 2 and no 3[1, 16].
4. 1. 1 The classical « size
effect » approach.
-For better
understanding
the transport mechanismsas a function of the films
thickness,
acomparison
ofthe
experimental
results with thepredictions
of thegeometrical scattering theories, namely,
F-Stheory [21]
and M-S
theory [22]
has been made. It isworth
to note that the F-Stheory
canonly
beapplied
to mate-rials with
spherical
energy surface.Nevertheless,
thecase of an
ellipsoidal
energy surface has been formu- latedby
P. J. Price[23]
and itsapplication
to bismuthfilms was done
by
A. N. Friedman et al.[24]
and M.Subotowicz et al.
[25]
who observed that the results obtainedby using Price’s
modelonly
differ a fewpercent from those obtained with the F-S
theory.
The relative
change
of theconductivity 03C3F/03C30
withthickness,
where ao, is the bulkconductivity that
hasbeen assumed to be
at 300 K and 77 K
respectively [26]
and 6F is the filmconductivity,
wasanalysed
within the classical size effectapproach by
Fuchs[27]
and Sondheimer[21] ]
and for two extreme cases the
following expressions
were found
[21] ]
for K >
1,
where K = andfor K 1.
Where d is the film
thicknesse, 1.
the mean freepath
of the bulk bismuth and P the reflection coefficient of the carriers at the film surface.
Figures
7a and b show theexperimental
resultscorresponding
to process nol, 2
and 3 for the thicknessdependence
of theconductivity
ratioUFlao at
300 Kand 77 K. The
comparison
betweentheory
and expe- riments has been carried out with values of10 ranging
from 1 000
A
to 3 500A
forexperiments
at 300 K[28]
and from 35 000
A
to 50 000A
forexperiments
at77 K [18].
From
figures
7aand b,
it can be concluded that the best fit is obtained for measurements made at roomtemperature
and for filmsprepared according
toprocess n° 2. The deviation from the
theory
is moremarked at 77 K for the other two processes.
This
discrepancy
cannot beexplained by giving
Pa
higher
value which in any case wouldsimply
cometo
give 10
stillhigher
values. Itis, by
the way, veryinteresting
to note that thevalue, 10
= 1 400A,
which
gives
the best fit at 300 K for process nD 2 is consistent with a conduction processthrough grains
of dimension
ranging
between 1 000A
and 2 000A
as was
directly
observed in thepreceding
SEMobservations.
1
Neverthless,
the F-Stheory
does not take into ,account the
scattering
of carriersby grain boundary
surfaces andimperfections,
and the sizedependence
of the
transport properties
can besignificantly
modi-fied
by
the presence of thesescattering
processes[29].
The
problem
ofscattering
of carriers atgrain boundary
surfaces was
carefully investigated by Mayadas
andShatzkes
[22] by including
thechange
ofpotential
atgrain boundaries ; they
have found the ratio of theconductivity
6g of thepolycrystalliné sample
to theconductivity
Qo of thesingle crystal
for a 0 to beindependent
of film thickness and to begiven by :
where,
dg
is the linear dimension of thegrain
and R theprobability
for a carrier to be reflected at agrain .boundary (0 R 1).
Using
the results ofMayadas
andShatzkes,
Molaet al.
[30]
have found thefollowing approximate expression
for the ratioUF/UO in
the presence of thegrain
boundaries and for the range0.2 K
5 :which have the
advantage
ofbeing
thicknessdepen-
dent.
It is clear that the
expression
of Mola et al.only
differs from the one obtained
by
F-Sby
the factorFig. 7a. - Thickness dependence of the room temperature conduc- tivity ratio for different processes. The continuous curves are
theoretical according to F-S theory for different mean free path
value : 03C30 = 7.35 x 105 MKS [14], P = 0.
Fig. 7b. - Thickness dependence of the conductivity ratio at 77 K
for different processes. The continuous curves are theoretical
according to F-S theory, for different mean free path value :
03C30 = 3.03 x 106 MKS [14], P = 0.
1 - R ;
and it ispossible
to obtain the samequality
of fit
by choosing
the values of10
and R as in table III.Table III.
Consequently
andtaking
into account the measured value of thegain
size(1 000 A-2
000Â),
it seems thatfor bismuth films
prepared according
to process n°2,
the presence of
grains
isconveniently represented by a
value ofthe mean freepath
of the orderof l0 ~ 1000 Á
and a very low reflection coefficient associated with the carrier
transport
between twograins (R 0.1).
Even
though
M-Stheory
includes thegrain
boun-dary scattering,
it fails to take into account thescattering
effectopposed by
defects such as surface states which areexpected
to be presentpredominantly
at the
grain boundary
surfaces for filmsprocessed according
to process n° 1. With the aid of the model hereafter thequantity
of surface states associated to this process will becomputed.
4.1. 2 The surface states model. - In what follows
a
quantitative study
of the surface states influenceon the film
conductivity
is outlined.Property
isaccording
to which U2 > a,, where thesubscripts
2and 1
respectively
stand for process n° 2 and n°1,
will be accounted for within the framework of this model.Indeed,
the structuralstudy
hasprovided
identical results for these two processes, as a conse- quence, the increase in the
conductivity
afterannealing
will be
entirely
attributed to the removal of a surface statesdensity.
In
considering’
the details of agrain-boundary scattering mechanism,
the idea of quantum mecha- nical tunnel has beengiven quantitative
and detailed examinationby
Drumheller[31]
whoshowed,
inparticular,
that the tunnel mechanism forgrain boundary penetration
canpredict
the films ohmic behaviour.The
theory [31] ]
leads to thefollowing expression
for the
conductivity :
which
provides,
in terms of the notation infigure 8,
theconductivity
value in03A9-1 .
m-1if m*,
the effectivemass of the
tunneling
carrier is inkg ; ô,
the effective thickness of the excesscharge layer
in thecrystallite
in
m ; d,
the film thickness in m ; a, theintergrain spacing in m ;
ç, the barrierheight
injoules
and Kthe relative
intergrain
dielectric constant.The numerical
application
has been carried outby assuming
a hole conduction mechanism and therefore m* = 0.16 x mo[32].
As in[31],
thefollowing
valueshave been chosen : k =
3, ô
= 3 x 10-’ andEF
= 0.23 eV. The most sensitive parameters in(5)
are a and ç that have been chosen to be ~2 = 1 eV and a = 7
A
toprovide
the most convenient fit toexperimental
values of U2 in[1].
Fig. 8.
- Energy band diagram at the grain boundary between two crystallites.For the value of 03C31 to be obtained with
[5]
it will beassumed,
in agreement with the structuralanalysis,
that all the
preceding
parameters remainunchanged excepted
(p which takes on alarger
value(çi > ~2),
as a result of the
larger density
of surface states inprocess n° 1
(Fig. 8),
which causes thesystematic
reduction for Q observed with unannealed films.
The ratio
03C32/03C31
is thengiven by :
This
equation
has been solvedgraphically
for03C32/03C31 = 2, in
agreement with(1)
and ~1 = 1.3 eV has been found.This result is consistent with the situation
depicted
in
figure
8 where the presence of adensity
of donor-like surface states
NS
causes the build upof compensat- ing
electrons at the surface of thecrystallite.
For theseelectrons to be
supplied
the energy bands bend at the surfacethereby causing
the barrierheight
to increase.For
NS
to becomputed
arelationship
between~1 - CP2 and
Ns
will have to be foundby solving
Poisson
equation
in thecrystallite
surfaceregion.
The evaluation of
dECV/dX
in S can then becomputed
in a way similar to
that_used
in semiconductor surface calculations[33]
with the result :where Eo = 8.85 x
10-12 F . m -1
andNe
the effec-tive
density
of states in the bismuth conductionband,
is
given by :
if the electron effective mass
density m:e
is taken to be0.05 x mo
[32]
the valueof NC
is 2.62 x1023 m-3.
For
computing
theintegral
in(7)
theassumption
of
complete degeneracy
has beenmade,
so that[34] :
the relation between
NS
and(EF - ECB)
then takeson a
quasi-linear particularly simple,
form :Therefore the surface state
density co-rresponding
tothe films of process n° 1 will be
given by :
and for ~1 - ~2 = 0.3
eV, (11) gives
This
is a very reasonable number which is consistent with theevaporation
condition of the films(50 Â/s
and 10-6
torr) [1].
Asatisfactory
linearapproximation
of
(11)
within the range :appears to be
given by :
which allows to derive a
particularly simple
and usefulrelationship
betweenNS
and 03C31. In the case of bismuth(~2 = 1 eV, a = 7 À
and m* = 0.16x mo)
thisexpression
is :4.2 HALL CONSTANT
RH.
-Figure
9 summarizessome results
obtained
for the Hall Constant value as afunction of
temperature
for unannealed films(process
n°
1) deposited
on different types of substrates and from different authors[12, 13].
Ageneral
feature ofthese results is the
change
in thesign
ofRH
for tempe-ratures
ranging
between 100 K to 200 K.Fig. 9. - Variation of the Hall constant versus temperature for Bi films prepared according to process n° 1. According to (2), glass
was used as substrate ; According to (3), glass was used as substrate ; According to (1) resin was used as substrate.
Figures
10 and 11 also summarize resultscorres- ponding
either to filmsincluding
a thermal treatment afterdeposition (process
n°2, Fig. 10)
or to thosegrown on hot substrates
(process
n°3, Fig. 11).
Theirmost salient
properties
are :i)
The value ofRH
remainspositive
for the whole range oftemperatures examined.
ii)
The values ofRH particularly
at lowertempe-
ratures are
highly dependent
on the film thickness andcan
attain
values 10 timeslarger
than those in processno 1 for the thinner films.
Figure
10 showsthat
our earlier results[1]
can becontinuously
extended into the thicker film domainby including those
of Ivanovet al.
[14].
This extension tends to show that the pro-perties of RH
do not seem to besignifiçantly
influencedby
the substrate material on the one hand and thatfor very
thick films(15 000 Á)
the effect of theannealing
process, if
measured
in terms ofRH,
ispractically destroyed
on the other hand.Fig. 10. - Variation of the Hall Constant versus temperature for Bi films prepared according to process no 2. Continuous lines corres-
pond to data obtained by (4), mica was used as substrate with
annealing temperature 240 OC. Other curves as in (1) resin was used
as substrate with annealing temperature 240 °C.
The
change
in thesign
ofRH
at lowtemperature
has been observed and firstinterpreted by
Le Traonet al.
[4]
in terms of structuralchanges, namely,
thequality
of the texture would beaffected by
the filmFig. 11. - Variation of the Hall Constant versus temperature for Bi films prepared according to process n° 3. Continuous lines cor-
respond to data obtained by (2), glass was used as substrate, substrate temperature 150 °C. Other curves as in (1), resin was used as
substrate, the temperature of the substrate was 220 °C.
thickness which is in marked contrast with the main conclusion of the structural
study.
Morerecently, however,
Inoue et al.[12, 15]
and Kochowski et al.[13]
have
interpreted
this resultby
the presence of surface states formedduring
condensation of the films. Therespective
values ofny’
andp03BC2h
govem thesign
ofRH
in :where the
subscripts
e and h refer to the electrons and holesrespectively ; n
and p aretheir
concentrationsand y
theirmobility.
However,
eventhough [14]
does notapply
toanisotropic
materials such as bismuth it has the greatadvantage
ofshowing
how sensitive the value and thesign
ofRH
can be to the influence of a surface statedensity. Indeed,
the evaluation of the variation of n and p at thegrain boundary
can bedirectly
derivedfrom the
preceding surface
states model.The
surface
states model also accounts forproperties i)
andii) ;
it has been shown that the presence of donor-type
surface state causes a decrease of the hole current contribution to the value of theconductivity
whichis due to the
higher
energy barrier related to the valenceband
(~1). Therefore,
it islikely
that the barrierassociated
to the free electrons in the conduction band is lowered andthereby
the electron current contri-bution can be enhanced. This effect accounts for the
change
in thesign
ofRH
at lowtemperature
for unannealed films. Moreover propertyii)
can then beinterpreted
asfollows :
surface states introducedduring
thegrowth
of thicker films are removed lesseasily
than for thinner onesby
the sameannealing
treatment. This is in agreement with the result of
figure
10 where the annealed thick films(15
000A)
appear to
display
a behaviour inRH(T)
which istypical
of unannealed films.In conclusion because of the
important
roleplayed by
surface states, the sizedependence
of thetransport properties
aredrastically
altered for bismuth filmsprepared according
toprocess
n° 1 where theirdensity
has been evaluated to be of the order
of 2 1012 cm-2
and hence the deviation ofexperimental
curves fromthose
predicted by
F-S and M-S theories is accounted for.As stated above the low value
(0.05
xmo)
of theelectron effective mass in bismuth has a great influence
on the
sensitivity
of itsconductivity properties
to thepresence of a surface state
density
at thegrain
surface.Films
processed according
to process n° 3 have shown to be made upby polycristalline grains superimposed
to an array of textured
crystallites,
the size of the disorderedgrains increasing
with thickness. This result isperfectly
consistent with theexperimental
observation
according
to which themobility
of thesefilms decreases very
rapidly
with their thickness[1].
The
large
disorientedpolycrystals
are thelimiting
agent of the conduction process
mainly
for thick films and for this reason no agreement between thepredic-
tions of either the F-S or the M-S theories and the
experimental
results should either beexpected
forthese films.
It has been stated that surface states will cause the energy at the
grain
surface to rise. The temperaturedependence of
theconductivity
and inparticular
thevalue of
Tc
at which a maximumconductivity
occurs, wheredQ/dt
=0,
may be discussed in terms of theprobability
ofovercoming
these barriers.Indeed,
thisprobability
increases withincreasing temperature
whereas at T >
Tc
alarger
role should be attributed toscattering
fromphonons leading
to a metallicbehaviour. It was
shown, indeed, [1 ] that
the maximumconductivity temperature
for filmsprepared according
to process n° 1 with the
highest
barrierheight (~1
= 1.3eV),
was not obtained within the rageTc
300 K while for filmsprocessed according
toprocess n° 2
(~2
= 1eV)
the value ofTc
was attainedwithin that range.
Appendix.
- The bismuth films obtainedby
ther-.mal
evaporation
onamorphous
substratekept
atroom temperature
(293 K) display
a texture such thatplane (111)rhombohedric
or(000 1) hexagonal
isparallel
tothe substrate. If we consider one of the
single crystals making
up the texture, itsreciprocal lattice
is apoint
lattice.
Figure
la shows everypoint
with the sameintensity.
A more strictfigure
should modulate thepoint intensity
as a function of the structure factor.The
reciprocal
lattice around the OZ* direction of the fibre axis. Each C*reciprocal lattice plane
consistsof a set of concentric circles
[35, 36].
It the electron beam is
perpendicular
to the surface of thefilms,
thecorresponding
diffractiondiagram
is made up of concentric circles and differs from the
Debye
Scherrerdiagram
of thepolycrystal by
theabsence of certain
rings. Indeed, only
remain therings (hkl),
such that their indices are related to those of the fibre axis(HKL) by
the relation :i.e. for the fibre axis
(OOOL),
therings (1120), (0330),
(2240), (1450),
etc...(the
indices are hererepresented
in
hexagonal notation).
If the
sample
is tilted withrespect
to thebeam,
the diffractiondiagram corresponds
to the intersection of thesphere
of reflection and thereciprocal
latticecircles.
A
particularly interesting
case is that forwhich
thenormal to the
sample
is tilted about 900 withrespect
to the electronic beam i.e. when the
sample
is examinedby
reflection. The OZ* axis is then in theplane
tangentto the
sphere
of reflection. On the theoreticaldiagram, figure 2b,
the diffraction spots will be found onthe Debye
Scherrerrings
at the intersection of theplanes C*,
2C*,
3C*,
... and thegenerant
ofcylinders reclining
on the circle lattice.Fig. lA. - Reciprocal lattice of bismuth.