• Aucun résultat trouvé

The relation between contact potential and planar conduction as a-Si : H films undergo gas adsorption or temperature changes

N/A
N/A
Protected

Academic year: 2021

Partager "The relation between contact potential and planar conduction as a-Si : H films undergo gas adsorption or temperature changes"

Copied!
12
0
0

Texte intégral

(1)

HAL Id: jpa-00209683

https://hal.archives-ouvertes.fr/jpa-00209683

Submitted on 1 Jan 1983

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

The relation between contact potential and planar conduction as a-Si : H films undergo gas adsorption or

temperature changes

J. Abelson, G. de Rosny

To cite this version:

J. Abelson, G. de Rosny. The relation between contact potential and planar conduction as a-Si : H films undergo gas adsorption or temperature changes. Journal de Physique, 1983, 44 (8), pp.993-1003.

�10.1051/jphys:01983004408099300�. �jpa-00209683�

(2)

The relation between contact potential and planar conduction as a-Si : H films undergo gas adsorption or temperature changes

J. Abelson

*

and G. de Rosny

Equipe Synthèse de Couches Minces pour l’Energétique, LPNHE, Ecole Polytechnique, 91128 Palaiseau Cedex, France

(Reçu le 15 février 1983, révisé le 18 avril, accepté le 28 avril 1983)

Résumé.

2014

Quand la surface libre d’une couche mince de silicium amorphe est exposée à de la vapeur d’eau, le potentiel de contact et la conduction planaire varient simultanément. Nous corrélons les deux effets

sur

trois échantillons et nous interprétons les

mesures en

terme de modifications, induites par l’adsorption, du potentiel

de surface, de manière similaire

aux

analyses des expériences d’effet de champ. Pendant les cycles thermiques,

on

observe

une

forte corrélation entre les portions

non

linéaires de la fonction liant le logarithme de la conduction à

1/T et la variation du potentiel de contact

Abstract.

2014

When the free surface of an amorphous silicon thin film is exposed to water vapour, the contact poten- tial and planar conductance simultaneously change. For three samples,

we

correlate the two measurements and

interpret them in terms of adsorbate induced changes of the surface potential, similar to field effect experimental analysis. During temperature cycles, there is

a

strong correlation between non-linear portions of the log conduc-

tance

versus

1/T plots and the variation of the contact potential.

Classification Physics Abstracts

72.80N - 73.30

-

73.60

1. Introduction

The planar conductance of a-Si : H may change by

orders of magnitude upon adsorption of certain spe- cies. Fitsche and Tanielian [1, 2] attribute this effect to a donor or acceptor like behaviour of the adsorbed molecules which accumulate or deplete carriers in a

conduction channel beneath the surface. However

they do not quantify the action of adsorbed species by measuring the induced surface charge density or

surface potential. Here, by simultaneously monitoring

the change in contact potential and planar conduc-

tance during gas adsorption, we demonstrate that the effect is similar to what is observed in field-effect

experiments [2-5], which supports the interpretation

of the above authors. The a-Si : H surface was expos- ed to H20 and Br vapour, water acting as an electron

donor and Br2 as an electron acceptor. The adsorption experiments were performed at room temperature

on samples either simply heat-dried or heat-dried and then illuminated by strong white light to induce

Staebler-Wronski effects [8]. We show that, on our samples, the difference between the two sample pre-

*

Present address : Material Science Department, Stan-

ford University, Stanford, CA. 94305, U.S.A.

parations originates partly from surface or interface modifications and that volume effects are also pre- sent

The determination of activation energies in a-Si : H

films is sometimes problematic as the logarithm of

conductance versus 1/T plots do not always exhibit unique slopes and furthermore, even in the favourable situations where an activation energy may be measur-

ed, the preexponential factor exhibits undesirable variations of several orders of magnitude. The origin

of these observations has been attributed to interface effects [5] or to more complex phenomena [7]. We

show that measuring contact potential as well as

conductance during thermal cycles enables one to

know if surface effects are present and whether the deduced activation energy and preexponential factor

are reliable in each specific case.

2. ExperimentaL

The experimental arrangement is shown schematically

in figure 1. The planar conduction of a-Si : H takes

place across a « gap » in the contact metallization of width d and height h

=

2.5 cm. The applied electric

field was smaller than 35 V/cm and ohmic conduction

was verified to take place. The sample contact poten- tial is measured by a Kelvin probe located over the

Article published online by EDP Sciences and available at http://dx.doi.org/10.1051/jphys:01983004408099300

(3)

Fig. 1.

-

Schematic of the experimental set up.

current collecting electrode. The substrate is mount- ed on a copper heater block. A vacuum enclosure

permits evacuation by a mechanical pump to 10 mtorr.

The adsorbate gases H20 or Br2 may be injected into

the chamber by a metered flow of argon carrier gas which bubbles through their liquids. A viewport

allows illumination of the sample.

The Kelvin probe or vibrating capacitor method

functions on the basis of establishing an a.c. null [9, 10]. The difference in work function between the Au

grid reference electrode and a-Si : H surface creates

an electric field between the two, which will induce

an a.c. current if the spacing and thus the capacity is

modulated by vibration. The applied d.c. level V c

is that needed to cancel the work function difference, making the a.c. current disappear : q. V, = OAu - Øa-Si. Thus an increase of Øa-Si as the bands bend up at the semi-conductor surface, causes a decrease in Vc

and vice versa. If the a-Si is not in thermodynamic equilibrium so that its Fermi level and therefore its work function cannot be defined, the a.c. null condi- tion implies that the following relation is verified :

q. Vc

=

’PAu - X. + qvs where X. is the a-Si surface electron affinity and vs is the potential of the conduction band edge at the surface, evaluated with respect to the back electrode Fermi level. Of course this last relation holds when thermodynamic equilibrium in the a-Si film is achieved for, in this situation, Oa-si

=

X. - qvs.

On gas adsorption and/or temperature changes,

the quantities oAu, Xr and vs may vary. We studied the

q5Au variations by evaporating 1 000 A of gold onto

the probe and a polished stainless steel plate, mounting

the plate in place of the a-Si and monitoring V c

for different treatments of the two surfaces. At the

beginning Vc was found null, then we rotated the

Kelvin probe away from the substrate and annealed the Au coated plate to 160° for 1 hour in N2. When the probe was brought into position Vc was equal to

200 mV, a reasonable difference considering that l/J Au

can vary by 1 V [11]. In one hour Fc equaled 120 mV,

as the probe heated over the hot substrate. One atmo-

sphere of water saturated air was admitted, producing Vc

=

12 mV 15 min. later. Pumping out the atmo- sphere yielded Vc

=

80 mV. Then we repeated the procedure of checking the evolution .in Vc when the

room temperature probe was positioned over the

hot Au plate. In vacuum no further evolution was

found as the temperature changed between the Au surfaces. This indicates that the probe work function had been stabilized by the successive thermal cycles.

The adsorption of H20 in saturated atmosphere on

the cold probe, but presumably not on the hot Au plate, decreased Vc by = 30 mV and was reversible.

Three a-Si : H samples have been studied : one obtained from an RF glow discharge of pure silane

provided by Proí Solomon, at Ecole Polytechnique,

one obtained under similar conditions at the Solar

Energy Research Institute (SERI) and the last one

prepared at SERI from a silane BF3 mixture. For

detailed parameters see table I.

3. Results.

3.1 GAS ADSORPTION. - We did not try to repeat the detailed study of Tanielian [2] on the effects of exposures of the a-Si surface to H20 partial pressures

as a function of pressure and time. Instead we used the gas as a tool to vary simultaneously the contact potential and the planar electrical conductance, seek- ing correlations between these two quantities. The complex chemistry relating the variation of the semi- conductor surface charge to the presence of a given

amount of water in the gas phase is then in some way

bypassed Figures 2 to 7 show the effect of water

adsorption on the these samples. Before the gas injec-

tion the samples had been either annealed to 160 OC

(state A) or annealed and then illuminated (state B).

The correlation between the contact potential V c

and the electrical conductance G is demonstrated by

the figures, each sample behaving in a specific way

which will be analysed in more details. We must note that the correlation between V c - and G is not strictly unique for a given sample, as can be seen on figures 6 or 7 where water desorption was initiated

Table I.

-

Sample characteristics.

(4)

Fig. 2.

-

Sample I : conductance

versus

contact potential

under exposure to H20 vapour, for

an

annealed initial state (state A). The time elapsed is quoted in minutes. The

curve

is

a

fit to data

as

explained in the text

by gas evacuation. The adsorption and desorption

curves do not coincide. However shapes are similar

and at a given G the discrepancy in Fc is at most

50 mV. Possible explanations for this effect will be discussed later.

The sample properties are analysed, using the following simplifying assumptions :

-

The adsorbed species act only on the film sur-

face and do not initiate any change in the bulk struc- ture which is furthermore assumed homogeneous.

They affect the film electrical properties only via

the induced surface charge or potential.

- The film electrical properties are however also controlled by the a-Si substrate interface charges or potential. The studied films are assumed thick enough

so that there exists a flat band region in the bulk.

Fig. 3.

-

Sample I :

same

data

as

for figure 2

on a

semi-log

scale. The straight lines slopes s

are

such that

The widths of the space charge regions may be estimat- ed from formula 6 of reference 6 : for a 1017 cm- 3 eV-1 uniform density of states in the gap, the space charge Debye length is 8.5 x 10-2 Jlm. The films under study are thick enough to insure the exis- tence of a flat band region, provided that the density

of states is not much smaller than the above value.

In this situation, the interfacial region is not affected by the modifications arising at the free surface. The observed conductance changes may be attributed to

changes only in the free surface space-charge region.

The surface potential VS may then be referenced with respect to the flat band region : flat bands at the sur-

face imply Vs

=

0 by definition.

-

A quantitative analysis similar to the one per- formed in field-effect experiments is possible at the

expense of the same assumptions discussed by Good-

man and Fritzsche [4]. Furthermore, in the present

case, the observed contact potential variations may arise from a variation of both the a-Si affinity X.

and the surface potential V. (the OAu variation was

shown to be small and is neglected). We shall assume

that the variation of X. is negligible and identify the V,

variations with the variations of the surface potential.

The validity of this assumption is not evident as H20

is polar and Helmoltz dipoles may easily build up at the surface, modifying X. [12]. The resulting rela-

tions between the conductance and the surface poten- tial, assuming 0 K statistics, are developed in the appendix. They involve a few unknown parameters that can be adjusted to fit with the experimental data.

See the appendix for the definition of the conduc- tance used in the following analysis.

Sample I.

-

In figure 2, the initial conductance decrease is followed by an increase, while the contact potential increases monototically. This behaviour reflects that the bands, initially curved upward at the surface, are progressively lowered and are finally

curved downward : H20 is acting as an electron donor and the carriers are initially holes and finally elec-

trons.

For large band bendings, the conductance is domi- nated by the transport in the region close to the surface

[4, 6]. It varies exponentially with the surface potential

if the density of states is constant in the gap region swept by the Fermi level : G oc exp ± qVs/kT [6].

Figure 3 shows the same data as figure 2 but on a semi- log scale, and one verifies that at large G the plot is

linear. The slope at small Vs is

-

q/kT however

at large Vs the slope is 0.5 q/kT. Such a behaviour is

expected if the density of states increases exponen-

tially as a function of energy around the flat band E-E

Fermi level: N(E) - N(EF) exp E y F where N(EF)

is the density of states at the flat band Fermi level

(formula A. 4). From the observed slopes of Ln G/ VS

one deduces that EY N kT

=

26 meV (see for-

mula A. 4 and A. 4’ of the appendix). This slope

(5)

Fig. 4.

-

Sample I : conductance

versus

contact potential

under exposure to H20 vapour, for

an

illuminated initial state (state B). The time elapsed is quoted in minutes.

is somewhat larger than the ones deduced from field effect experiments [4, 5]. Moreover the 0 K statistics

approximation is poor in this situation.

One can try to describe the whole curve using for-

mula A. 3 of the appendix :

G(Vs) is the conductance, Vs the surface potential

referenced to the flat band position. G(0) is the con-

ductance for flat band conditions at the free surface,

it includes a possible contribution due to band bend-

ing at the silicon-substrate interface. In and Ip are

mathematical functions depending on Vs and E.,

as detailed in the appendix, N(EF) is the density of

states at the Fermi level in the bulk region, Qn and up

are the bulk conductivities of electrons and holes

respectively.

The conductance is minimum at a value of the

n’

m

h

qvm

surface potential VS satisfying the relation : kT

=

Ln P (A.2) (this holds for any density of states an

distribution). The experiment provides the value V’

of the contact potential at which G is minimum. Vr

has been related to V. assuming that these potentials

are shifted by a constant quantity Y° so that Vs

=

Vc-

Vf, yo is the contact potential for flat band condi- tion at the free surface.

In our experiment, Vf is deduced from a fit to the whole G(yS) function using formula A.3.

(Ey

=

26 meV is used as discussed above). Some

authors [13] use the hypothesis that under strong illu- mination the bands should flatten thus providing directly V0c from a Kelvin probe measurement under illumination. The results we obtained by this method

were not reliable so we discarded its

The fit (curve of Fig. 2) yields the following values :

-

The flat band contact potential Y° is 280 mV

and the value of the conductance G(0) is 4.5 x 10-140-1.

-

V7, the value of contact potential at the con-

ductance minimum is 175 mV, yielding Vm

=

-

105 mV. From (A. 2), one gets Gp/Gn

=

0.02.

This implies that the bulk of the sample under study

is slightly n-type. More precisely the Fermi-level is located - 50 meV above the position which would lead to an equal contribution of electrons and of holes to the conductivity.

Note that in contrast to conventional field effect

experiments, the up/an ratio is obtained without the need of any extra information. See reference 4 for a discussion on this subject.

An upper limit for an may be deduced as follows : when the bands are flat at the free surface, the con-

ductance G(0) is the sum of the flat band conduc- tance GFB in the whole sample and of the conduc- tance at the semi-conductor substrate interface GI.

The latter may be expressed as a function of the inter- face potential by the same formula as the one esta-

blished for the conductance at the free surface, since the sample has been assumed homogeneous. It is

then easy to show GFB 2 G(0) - Gm where Gm =

2.8 x 10-14 Q-1 is the value of the conductance

minimum, so that GFB 6.2 x lO-14 Cl - ’. Now

GFB

=

e( Qn + Qp) N eGn where e, the sample thick-

ness is 0.5 pm, so a. 1.2 x 10- 9 0-1 cm-1.

-

uan it 1 c F’o is found

e ual to

-

The

quantity - V N(EJ Un is found equal to

q N(EF)

n

1.1 x 10-14 0-1. One then deduces an upper limit for N(EF), the Fermi level density of states :

This is a typical value [6].

Figures 4 and 5 show the conductance dependence

with contact potential of sample I initially prepared

Fig. 5.

-

Sample I : same data

as

for figure 4 on

a

semi-log

scale. The straight lines slopes

are

such that

(6)

in state B (Staebler-Wronski), in linear and semi-log plots respectively. The initial conductance is slightly

lowered compared to state A and the shapes of the

curves are not the same. The slopes at large G (Fig. 5)

now suggest a flat density of states, however one does not find any satisfying fit to the data. Light induced

thickness inhomogeneity is a possible origin of the discrepancies between states A and B.

Sample II.

-

Figure 6 shows the conductance as a

function of contact potential under exposure to H20

vapour. Curve A corresponds to an annealed initial state (state A) whereas curve B corresponds to an

illuminated initial state (state B). Curve B has been

shifted upward by the amount indicated by the arrow.

The two are approximately superposable, although

state B exhibits a slightly higher initial conductance. A noticeable feature is the lower asymptotic conduc-

tance and contact potential induced by water on

state B as compared to state A. This effect has been observed by Tanielian [2], who uses it as an argument in favour of light induced bulk defects. On the present

sample, the similarity of the two curves suggests rather a light induced surface modification that limits the H20 activity.

The constant conductance increase with contact

potential indicates that the bands are initially either

flat or curved downward, however if aplan 1 they

may also be slightly curved upward At large G, the log G/ VS slope is too small to be compatible with the

conductance model in the appendix.

Fig. 6.

-

Sample II : conductance

versus

contact potential

under exposure to H20 vapour, elapsed time in minutes.

Curve A : annealed initial state (state A). Section 0-11 : exposure to H20, sections 16-20 and 0-28 : effect of pump-

ing. These two sections

are

separated by

an

unrecorded time.

Curve B : illuminated initial state (state B). For clarity the

data have been shifted upward by the amount indicated by

the

arrow.

Figure 7 shows the correlation between conduc- tance and contact potential under exposure to Br vapour. The evolution is very fast and difficult to

control, so the data are of poor quality. The conduc- tance increases by two orders of magnitude while the

contact potential decreases by - 300 mV : Br acts as

an electron acceptor and the bands curve upward. The

action of Br on the probe has not been verified, so there may be also an appreciable variation of the gold

work function under exposure to Br.

From figures 6 and 7 ones deduces that the conduc- tance minimum Vm lies between V c

=

150 and 250 mV.

It appears that, within the caveats on the interpretation

of figure 7, G increases more slowly as a function of

Vc when V, increases from V7 than when it decreases.

This is similar to the situation displayed on figure 2

and it indicates that up/ an 1 and that the flat band

position Vf is higher than Ym. The lack of a reliable model prevents however a quantitative determination

of Vco.

From the lowest measured conductance - 2 x

lO-14 Q-1 and using the arguments developed for sample I, one gets an upper limit for an of the same order as for sample I : an 10-9 Q-1 cm-1. No quantitative evaluation of the density of state is reliably achieveable, however the large conductance sensitivity

to the contact potential variations indicates that it must be around or less than 1017 cm-3 eV-1. It may indeed be so low that there is no flat band region in the sample. Then changes in surface potential would

induce changes at the interface which would explain

the failure of the conductance model developed in the appendix. This interpretation should be ckecked

using a thicker sample.

Sample III. - This sample was prepared with a 3 % BF3 in SiH4 mixture so- it is presumably p-doped Figure 8 shows the change in conductance as a function of contact potential when the sample is exposed to H20

vapour. Curve A corresponds to an annealed initial state (state A), curve B to an illuminated initial state

(state B). There is a net shift of the conductance

Fig. 7.

-

Sample II : conductance

versus

contact potential

under exposure to Br, vapour, for

an

annealed initial state

(state A). The time elasped is quoted in minutes.

(7)

Fig. 8.

-

Sample III : conductance

versus

contact potential

under exposure to H20 vapour. Curve A : annealed initial state (state A). The time elapsed is quoted in minutes. The

curve

is

a

fit to data

as

explained in the text Curve B : illuminated initial state (state B). Section 0-20 exposure to

H20- Section 20-120 : effect of pumping.

between the two that may be attributed to an illumi- nation induced change of the substrate interface

potential. Moreover they do not exhibit the same

curvatures, which indicates a modification of bulk states. The initial and final contact potentials are only

shifted by N 50 mV between the two curves, suggesting

a minor effect of illumination on the action of H20 at

the sample surface.

The conductance decrease when Vr increases indi- cates an initial hole transport This decrease is slower than exponential and tends to a minimum. This indi- cates that the flat band condition is in the vicinity of the

initial V c value and minimum conductance not far

beyond the final Vc value. Indeed curve A may be

fairly well described using a flat band approximation (formula A. 5, curve on Fig. 8); Vm was fixed at

650 mV and Vc0 is found at 200 mV.

Using the same arguments as in the study of sample I,

one finds a ratio Qp/ Qn

=

3.3 x 10’ at flat band, while

the Fermi level is located - 225 mV under the position leading to equal electron and hole conductivities. This demonstrates that the sample is slightly p-doped

From distribution A one gets a up upper limit of the

order of 2.5 x 107 g-1 cm-1. The quantity

1 880 - . cr is found equal to 1.5 x io-12 c)-l

and the corresponding upper limit for N(EF) is then

1.6 x 1017 CM-3 eV- 1.

3. 2 THERMAL CYCLES - After exposure of the sample

to H20 vapour the enclosure is pumped out, then the conductance and contact potential changes are moni-

tored during temperature cycles. Figures 9 to 11 show

their variations as a function of 103/T for samples I

to III respectively. Each figure shows the effect of the first heating and cooling, which differ because of the water desorption during heating. On subsequent cycles, either heating or cooling induce G and

variations that are close, but not identical, to the first cooling cycle.

The effects of surface and interface potentials on the

T dependence of the conductance have often been stressed [5, 6]. In the present experiment, measurement of the contact potential brings some information on

the surface, but contrary to the fixed temperature adsorption experiments, the effect of the interface cannot be separated

Sample I (Fig. 9).

-

The values swept by the contact potential show, by comparison with figure 2, that the surface is highly p-type during the thermal cycles. If

the interface potential does not vary too much during

the experiment, one may infer that most of the conduc- tance takes place at the free surface. Indeed one observes on figure 9 that at a given T, the higher

conductance corresponds to the lower potential as expected. Moreover, in the surface potential region

under consideration, one expects that the following

relation between the conductance and the surface

potential holds at fixed T : Ln G

= -

q Vs/k T + C,

where C does not depend upon V. (formula A. 4 of the appendix). This formula is used to find the expected Vc

upper branch, shown in open circles, using at each T

the measured conductances and the lower branch value of Vr. The origins of the obvious discrepancy of

Fig. 9.

-

Sample I : contact potential (upper figure) and log conductance (lower figure)

versus

103/T. The

arrows

indicate the direction of evolution from

room

temperature.

Before the thermal cycle, the surface had been exposed to

water vapour and then put under vacuum condition (- 10- 2 torr). Circles : position of the expected V c upper values computed from conductance and lower Vc measure-

ments

as

explained in the text

(8)

the order of 70 mV may be multiple : the variation of the gold work function resulting from water desorp-

tion could account for 30 mV or a sample too thin compared to the Debye length, so that no flat band region exists, would invalidate the above formula.

These possible effects are now under study.

Sample II (Fig 10). - Here again the variation of conductance with surface potential at fixed T is in qualitative agreement with the data of figure 6. Notice

the straight line in the Log G/( 1 / T) (activation energy 0.63 eV) plot associated with a constant surface poten- tial during the cooling cycle. The V, value - 1 20 mV

corresponds to a position close to the conductance minimum at room temperature (see Fig. 6). A straight

line is expected when the surface conductance domi- nates, which corresponds to a strong bending of the

bands. The measured activation energy is then

Eac - I e V s I [6] where E3£ is the bulk activation energy of the carrier type accumulated at the surface. It is however presumably not the situation encountered here since V. is close to the conductance minimum,

where both carriers contribute to the surface conduc- tance. More probably, one is in a situation where the free surface conductance is negligible compared to the

conductances in the bulk and/or at the substrate inter- face. A pure bulk conductance would imply a pre- factor ao

=

3.8 fl-’ cm’ which is too small and indi- cates an appreciable contribution of the conductance at the substrate interface [6]. The potential at the

interface would then be independent of T to account

for the observed straight line in the Log G/(1/T) plot

Sample III (Fig. 11).

-

The qualitative behaviour of t e Log G/(1/T) plot may again be understood using figure 8. Note that the Log G curve does not cross at

the position where the Vc curve crosses. This effect may be due to a strong modification at the interface, as already stressed when discussing figure 8. The Log GI(I IT) plot is difficult to interpret quantitatively

as it involves surface potentials where both holes and electrons contribute to the conduction. The last part of the cooling cycle is performed with an approximately

constant Vc ’" 200 mV, which correspond to a flat

band condition at the surface. The corresponding

activation energy is E.

=

0.3 eV, the pre-exponential

factor is G = 1.18 x 10-6 fl-’ and they yield

G

=

10- 11 Q - 1 at room temperature, a value compa- tible with the one found on figure 8. If one may neglect

the contribution of the substrate interface, the measur-

ed Ea is to be identified with Eap, the activation energy for holes. One then deduces that the level Eo corres- ponding to equal electron and hole conductivities is located 525 meV above the valence band conduction

edge, as it has been shown previously that the Fermi level is located 225 mV below Eo. Remembering that

the conductivity gap width is in the 1.6-1.8 eV range [6],

one sees Eo is much closer to the valence band than to the conduction band It implies that the hole mobility

is much lower than the electron mobility. Here again

one must take care of a possible underevaluation of the activation energy due to the conductance at the subs- trate interface, so that on this particular sample the

most reliable information comes from the adsorption experiment

Fig. 10. - Sample II : contact potential (upper figure)

and log conductance (lower figure)

versus

103/T. The arrows

indicate the direction of evolution from room temperature.

Before the thermal cycle, the surface had been exposed to

water vapour and then put under

vacuum

conditions (- 10-2 torr).

Fig. 11. - Sample III : contact potential (upper figure)

and log conductance (lower figure)

versus

103/T. The arrows, indicate the direction of evolution from

room

temperature.

Before the thermal cycle, the surface had been exposed to

water vapour and then put under vacuum conditions

(- 10 - 2 torr).

(9)

4. Discussion.

The assumptions used to analyse the data include several commonly applied to field effect experiments [4]. The objections raised to the latter are also pertinent

here. In particular, the strong sensitivity to a possible inhomogeneity close to the surface and the invalidity

of the conduction model under strong band bending

are often pointed out They throw some doubt on the

deduced Fermi level density of states. However other

methods sensitive to the bulk density of states give

values only slightly lower compared to the ones

deduced from field effect data [14]. Two extra hypo- theses, specific to the present experiments, are the

identification of the contact potential Vc with the

surface potential V. (given a constant offset) and the assumption that the gas is only adsorbed superficially

and does not diffuse into the sample. The observation of a non unique relation between Vc and G during

fixed temperature adsorption experiments may be due to the invalidity of one or of both of these assumptions.

It may also imply that the sample is not really in equili-

brium and that long time constant evolution takes

place under vapour exposure. A way to check this last

point would be to perform the same experiments while insuring a much slower evolution of the parameters.

The repartition of the contact potential between xS, the electron affinity and Vs the surface potential would be

accessible if it can be proved that strong illumination induces flat bands at the surface without changing

other properties of the material. Measuring Vc under

such a condition then allows one to determine xs [ 13].

Independent measurements of xs would require ultra high vacuum conditions which are not compatible

with the presence of vapour in the vessel.

The present experiment shows that the planar

conductance measurements are extremely sensitive to

the value of the surface potential as already stressed for instance in reference 6. Here we are able for the first time to demonstrate directly the effect In particular, it

is possible to have a dominant surface channel conduc- tion by holes in samples that are otherwise slightly n-type. These observations emphasize the care that

must be taken when interpreting planar conduction experiments. Similar difficulties are expected to arise

from the interface with the insulator. An experiment combining field effect technique and Kelvin probe

measurements would allow one to control both surface and interface potentials.

The surface potential is observed to change during

thermal cycles. This change may be correlated with a

variation of the slope of the Log G/(1/T) distributions.

In the temperature intervals where the contact poten- tial is observed to be approximately constant, the

corresponding slope is also constant It would however be meaningless to interpret the slope in terms of activa-

tion energy without evaluating and subtracting the

surface channel contribution to the conductance. This evaluation is in principle possible if one is able to interpret quantitatively the gas adsorption experi-

ments, however nothing is known about the interface

potential. Combination of Kelvin probe and field

effect would again allow control of both surface and interface potentials during thermal cycles.

Light soaking is shown to induce changes at the surface, the interface and in the bulk. The three samples

behave in distinct ways. For sample II there is a

reduction of the band bending variation under water

adsorption, which is interpreted in terms of light

modification at the surface. For samples I and III, the bulk appears to be modified, as evidenced by the modi-

fication of the relation between the contact potential

and the conductance. Moreover on sample III, there is also evidence for a light induced change of the inter-

face potential. The present results have to be contrasted to the ones obtained by Tanielian [2] who finds that the major effect of light illumination is a reduction of the band bending variation induced by gas adsorption.

His explanation of the effect by a modification of the bulk density of states is questionable given present

contact potential measurements.

Finally, the experiment is now being redone on

thicker samples to be sure of getting a flat band region

in the bulk, so that the modifications arising at the free

surface do not modify the potential profile at the

substrate interface.

5. Conclusion.

It has been shown on three samples that a measurement of contact potential as well as conductance during

both gas adsorption and thermal cycles may bring original information about the sample. The inter-

pretation of adsorption measurements is similar, but simpler, than in field effect experiments. Strong light

illumination modifies both the film surface and bulk

(Staebler-Wronski effect). The validity of activation energy measurement, often problematic due to surface

or interface effects may be checked if one simultaneous-

ly follows the contact potential variations. Moreover, in certain situations, the Log G/(1/T) distributions may possibly be corrected to yield the correct activa-

tion energy.

Acknowledgments.

The authors wish than Prof. I. Solomon for stimulating

discussions and for providing one sample, Dr. R. Kerns

of SERI for providing the two other samples, and

G. Benet for his skilled construction of the Kelvin

probe.

(10)

Appendix.

RELATION BETWEEN CONDUCTANCE AND SURFACE POTENTIAL.

-

The following approximations, frequently

used in field-effect or capacitance analysis [4] are assumed to be valid :

-

homogeneous material

-

activated electron and hole transport taking place at the respective mobility edges

-

conductivity prefactors independent of the band bending

-

zero temperature statistics.

Furthermore the sample is assumed thick enough to ensure the existence of a region where space charge

effects may be neglected (flat band or bulk region).

DEFINITION OF THE CONDUCTANCE.

-

Let d be the conductivity « gap » width (Fig. 1), h its height, V the applied potential across the « gap » and I the measured current The conductance G used in this article is defined as

G

=

I/V x dlh.

It is such that G

= u(x) dx, u being the sample conductivity at a distance x from the surface, e being the sample thickness.

RELATION BETWEEN G AND Vs.

-

6(x) is a function of x through V(x), the value of the potential at the distance

x (see Fig. 12) :

an and Qp are the bulk conductivities for electron and hole transport respectively Ean, Eap are the corresponding

activation energies and U.nl up the corresponding conductivity prefactors.

V(x) is related to the charge density and hence to the density of states by Poisson’s equation. For a given YS (the value of Y(x) at the surface) the solution V(x) is unique in the surface space charge region. Moreover, the assumed film homogeneity implies that the solution Y(x) corresponding to various V. are obtained by transla-

tions along x of a unique curve.

Let xl be a value of x lying in the flat band region then :

The second integral does not depend on V..

The first integral may be written :

is the solution such that V

=

Vs at the surface.

Now let Vs be incremented by AVs, and let åx be the position at which V(x, V

S

+ AVs) takes the initial value Vs. The translational property of the solutions implies that

Fig. 12.

-

Schematic of the electronic band curvature

as a

function of thickness.

(11)

so that

assuming small increments the preceding integrals may be written :

V(0, Vs) = VS by definition and V(xl, Vs)

=

0 as xl is in the bulk so

or

from (A. .1) one gets

so that

may be computed as follows : Poisson’s equation yields : , where p is the charge density

and N(E) is the density of states. This relation assumes 0 K statistics and that filled states under the Fermi level are

neutral and unfilled states above the Fermi level are also neutral. The important point here is that p depends on x only through V(x). This is also true for finite temperature statistics (Ref 5 formula 3). One may then easily perform a first integration of Poisson’s equation :

where

so

G is minimum at this value of Vs.

G(V.) may be expressed as follows :

(12)

the plus sign being associated with In and the minus sign with Ip :

These integrals may be evaluated numerically.

G(o) is the conductance for flat band condition at the surface.

When

then

when then then

1 one gets a constant density of states approximation, then

the + sign is associated to In and the minus sign to Icp.

References

[1] FRITZSCHE, H., Solar cells 2 (1980) 289.

[2] TANIELIAN, M., Philos. Mag. B 4 (1982) 435.

[3] MADEN, A., LE COMBER, P. G. and SPEAR, W. E.,

J. non-crystalline Solids 20 (1976) 239.

[4] GOODMAN, N. B. and FRITZSCHE, H., Philos. Mag. B

42 (1980) 149.

[5] GOODMAN, N. B., Philos. Mag. B 45 (1982) 407.

[6] SOLOMON, I., DIETL, T. J. and KAPLAN, D., J. Physique

39 (1978) 1241.

[7] FRITZSCHE, H. and TANIELIAN, M., Tetrahedrally

Bonded Amorphous Semi-Conductors edited by

R. A. Street, D. K. Biegelsen and J. C. Knights

AIP Conf. N° 73 (1981) p. 318.

[8] STAEBLER, D. L. and WRONSKI, G. R., App. Physics

Lett. 31 (1977) 292.

[9] MANY, A., GOLDSTEIN, Y. and GROVER, N. B., Semi- Conductor Surfaces (North Holland Amsterdam)

1965.

[10] BONNET, J., Thesis, Montpellier 1981, Unpublished.

[11] SURPLICE, N. A. and BREARLEY, W., Surf. Sci. 52 (1975) 62.

[12] MANY, A., Crit. Rev. Solid State Sci. 4 (1974) 515.

[13] GOLDSTEIN, B., DRESNER, J., SZOSTAK, D. J., Philos.

Mag. B 46 (1982) 63.

[14] MACKENZIE, K. B., LE COMBER, P. G., SPEAR, W. E.,

Philos. Mag. B 46 (1982) 377.

Références

Documents relatifs

The calculated ozone and temperature changes calcu- lated by the model are believed to represent upper limits (as- suming that no major reaction is missing in the chemical

The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.. L’archive ouverte pluridisciplinaire HAL, est

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des

(A) Effective surface viscosity as a function of the number of layer pairs i in the film for the same beads, and (B) evolution of the second derivative of the interaction potential V

• Pitch of harmonic complex tones does not change with sound

In the large temperature, non-degenerate limit (but still assuming the s-wave scattering regime), one can use the virial expansion (see Supplementary Note 4 and ref. 35 ) to

This present pilot study investigates the relationship between dependency distance and fre- quency based on the analysis of an English dependency treebank. The

— We prove that if a contact 3-manifold admits an open book de- composition of genus 0, a certain intersection pattern cannot appear in the homol- ogy of any of its minimal