• Aucun résultat trouvé

Passivation of Extended Defects in Silicon by Catalytically Dissociated Molecular Hydrogen

N/A
N/A
Protected

Academic year: 2021

Partager "Passivation of Extended Defects in Silicon by Catalytically Dissociated Molecular Hydrogen"

Copied!
8
0
0

Texte intégral

(1)

HAL Id: jpa-00249659

https://hal.archives-ouvertes.fr/jpa-00249659

Submitted on 1 Jan 1997

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.

Passivation of Extended Defects in Silicon by Catalytically Dissociated Molecular Hydrogen

S. Binetti, S. Basu, M. Acciarri, S. Pizzini

To cite this version:

S. Binetti, S. Basu, M. Acciarri, S. Pizzini. Passivation of Extended Defects in Silicon by Catalytically Dissociated Molecular Hydrogen. Journal de Physique III, EDP Sciences, 1997, 7 (7), pp.1487-1493.

�10.1051/jp3:1997201�. �jpa-00249659�

(2)

Passivation of Extended Defects in Silicon by Catalytically

Dissociated Molecular Hydrogen

S. Binetti (~,*), S. Basu (~,**), M. Acciarri (~) and S. Pizzini (~)

(~) INFM Istituto Nazionale di Fisica della Materia, Department of Materials Science, Via Emanueli 15, 20126 Milano, Italy

(~) Consorzio Milano Ricerche, Via Cicognara 7, 20133 Milano, Italy

(Received 3 October1996, accepted 14 March 1997)

PACS.66.30.Jt Diffusion, migration and displacement of impurities PACS.61.72 -y Defects and impurities in crystals; microstructure

Abstract. This paper reports the results of

a new

hydrogenation process, which applies the properties of noble metals

as

chemisorptive dissociation catalysts for molecular hydrogen. Used to passivate deep states in several kinds of polycrystalhne materials, H has been shown to be

particularly effective for samples grown by the EFG (Edge Film Grown) technique. These results

are

compared with former

ones

obtained

on

dislocated single crystals, which

were

passivated under

an

hydrogen plasma, to speculate about the role of dislocations

on

the yield of

a

hydrogen passivation process

1. Introduction

Atomic hydrogen from various sources is known to passivate surface states as well as shallow and

deep levels arising from dopants, impurities and defects in silicon or other semiconductors, thus

inducing important changes in their electronic properties, like the surface recombination rate, the ionized dopant profile, the minority carriers lifetime and the majority carrier concentration.

Albeit HF can be an effective but temporary surface passivating agent, permanent passivation is usually obtained using ion implantation or Radio-Frequency (RF) and microwave plasma

hydrogenation. Both these techniques present, however, an intrinsic disadvantage. In fact,

in both cases hydrogen atoms are physisorbed and chemisorbed at the semiconductor surface,

which is thus passivated and which behaves as an hydrogen reservoir continuously replenished

as long as the hydrogenation process continues, and from which H diffuses into the bulk until is trapped at a suitable point or extended defect. The effective passivation depth comes from

a delicate balance of inward and outward diffusion and of trapping and de-trapping processes, all having a strong temperature dependence. In addition, the diffusion of H in silicon itself is

known to be a complex process, influenced by quantum mechanical effects at low temperatures and by interactions of H with H itself or with other impurities at medium temperatures Iii. We expect therefore that the passivation depth and yield would strongly depend on temperature. if

we would be able to use extended defects like Grain Boundaries (GB) or segments of emerging (*) Author for correspondence

(**)Now at India Institute of Technology, 721302 Kharagpur, W. Bental, India

Q Les #ditions de Physique 1997

(3)

1488 JOURNAL DE PHYSIQUE III N°7

Table I. Imp~rity concentration of polycrystalline samples.

Samples [Cl (ppma) [O] (ppma)

EFG 15.6 ~ 0.5 2.2 ~ 0.5

EU 15.8 ~ 0.5 1.5 ~ 0.5

dislocations as preferential hydrogen chemisorption centers and as pathways for fast H-diffusion,

the further diffusion process to bulk H-traps would use as the atomic hydrogen source the extended defects themselves, thus strongly enhancing the kinetics of the overall hydrogenation

process and strongly increasing the in-depth passivated layer. Eventually, if we would be able to decorate selectively emerging dislocations and GB with a catalyst capable to favor the chemisorptive dissociation of molecular hydrogen, we would be able not only to use molecular

hydrogen instead of atomic hydrogen, but, possibly, to discriminate between different kinds of dislocations and GB, in view of their ability to be decorated with metals.

We will show that this process can be carried out using Pt as the catalyst on different types of polycrystalline and single crystal silicon substrates.

2. Experimental Details

2.I. SAMPLE SELECTION. Two kinds of polycrystalline samples, used as substrates for solar cells, were considered in detail, both exhibiting very low diffusion length (LD) values in

as-grown conditions and thus requiring passivation to improve the photovoltaic yield.

The former were grown by the "Edge Film Growth Technique" (EFG) [2] and chosen in view of their ability to show a large LD increase after plasma hydrogen passivation [3].

The latter, supplied by Eurosolare SpA (Nettuno, Roma), were grown by the directional solidification method [4] and cut from the lateral top region of a 55 x 55 x 21 cm3 multicrystalline

silicon ingot.

As a consequence of the different growth techniques used, the morphology, the microstructure and the impurity content of EFG and Eurosolare (EU) samples are very different. While their composition is reported in Table I, the dislocation density determined using the etch pit counting procedure is almost constant in EFG samples (around 10~ cm~2) and very variable from region to region in EU samples.

Moreover, as already shown in a previous paper [5], the EU samples are characterized by

the presence of silicon nitride and iron silicide submicrometric particles. The density of these

particles is very large in the edge region of EU ingots, where the minority carriers diffusion

length is remarkably lower than the ingot average value.

As reference, we used single crystalline samples, cut from high-oxygen ([Oi]= 22 ppma)

boron doped (p

=

1-2 Q cm) CZ-grown 4" diameter silicon wafers. As obtained by the

supplier, the wafers are dislocation free. These samples were then submitted to a dislocation

generation procedure which follows a practice established at the Laboratoire MATOP-CNRS

(Marseille). By this procedure [6] dislocation sources are nucleated by scratching the surface of the samples along the loll) direction with a diamond tip loaded with a 0.3 N weight (other

details are reported in [7]). The dislocation density, determined ma etch pit counting and

X-ray topography, is shown to vary from 10~-10~ cm~2 near the high stressed end of the scratch (maximum resolved shear stress 35 MPa) to 10~-10~ cm~2 near the low stressed end

(maximum resolved shear stress 4 MPa).

(4)

2.2. HYDROGENATION PROCEDURES. Many noble metals show catalytic properties for molecular hydrogen dissociation, including gold, platinum and palladium. In order to satisfy

both the chemistry and the physics of the hydrogen passivation process, however, the suitable catalyst should have a low diffusion coefficient in silicon, in order to avoid any electrical degra-

dation of the substrate associated to deep traps or recombination centers. For this reason, gold

must be rejected.

Moreover, as the dissociation of the dimer to atomic hydrogen can occur everywhere at the surface of the catalyst, H dissociatively chemisorbed at the catalyst surface should diffuse

quickly to the silicon-catalyst interface. Dissolution or retainment of hydrogen in the catalyst

could influence adversely the overall reaction kinetics. For these reasons, Pd must be rejected

in favor of Pt which presents the most favorable properties, including a very low diffusion depth

in silicon [8], at the hydrogenation process temperatures (T

=

300 °C, see below).

Pt films of various thickness were deposited on the silicon surface both by DC sputtering and by pyrolysing an aqueous solution of 7.72 x 10~3 M in (NH3)4Pt(N03)2 (9] at 250 °C, this last procedure being more suitable as creates a greater surface area for hydrogen dissociation.

Before the metal deposition the samples were chemically polished using a CP4 solution (HN03

HF

:

CH3COOH) to remove the cutting damage. The duration of the chemical treatment was

long enough to remove a 10 pm thick layer.

The catalytic hydrogen dissociation process was then carried out at 300

°

C, in a quartz tube, using a 5%H2/Ar mixture as the hydrogen carrier gas. The hydrogenation process lasts 2 h.

The furnace temperature and process time were chosen in order to have a temperature high enough to allow a fast diffusion of hydrogen in silicon, while keeping the hydrogen desorption

rate as low as possible, to minimize the diffusion of platinum in the bulk of the samples and to avoid the formation of the boron-hydrogen complexes which are stable up to 200

°

C [10].

In order to test possible lifetime degradation effects associated with Pt-diffusion during

the heat treatment, some preliminary experiments were carried out at 300

°

C, in a pure argon

atmosphere. The results of these experiments show that the electrical properties of the material

are practically unaffected by such treatment.

In order to compare the effect of catalytic hydrogenation with that obtained by using a plasma source, the same kinds of samples were exposed for 2 hours to a Radio-Frequency (13.56 MHz) driven hydrogen or deuterium plasma, using a Plasmalab system ii ii.

Before and after the hydrogenation process, all the samples were submitted to diffusion

length measurements using the SPV method, with a light spot of about IA x 4 mm2, in at least three different positions on each sample.

Before the SPV measurements and after the catalytic hydrogenation the Pt or land Pt sili- cides deposits eventually formed at the sample surface were removed by purple etch (10$lo HF: 48$loHN03~42%CH3COOH saturated with 12) at room temperature for 15 s [12].

As it has been demonstrated [13] that the effect of plasma hydrogenation is significantly

enhanced by a post-hydrogenation annealing at 450

°

C, which induces the'diifusion of hydrogen,

accumulated in shallow subsurface region (about 0.4 pm) towards the entire volume of the

material, we applied systematically this post-hydrogenation step both to samples which were

hydrogenated by RF plasma or catalytically.

Eventually, for microstructural studies before and after the hydrogenation procedures, Scan-

ning Electron Microscopy (SEM) was used on samples anisotropically etched in a Yang solution

(IHF:I(1.5MCr03)) which delineates defects on (iii), (100) and (l10) silicon planes.

(5)

1490 JOURNAL DE PHYSIQUE III N°7

Table III SPV meas~rements on EFG samples. LDI, LD2 and LD3 are, respecti~ely, the average dijfwion length of the as-grown, catalytically treated and post-annealed samples. On each sample,

a

minim~m of three meamrements in different position were carried o~t

Sample LDI (pm) LD2 (pm) LD3 (pm)

CS2 28 ~ 3 52 ~ 5 45 ~ 5

CG2 33~3 47~5 41~4

LS2 26~3 42~4 41~4

LG2 21~2 60~6 52~5

Table III. SPV meamrements on EFG samples. LDI, LD2 and LD3 are, respectively, the average dijfwion length of the as-grown, plasma treated and post-annealed samples. On each sample, a minim~m of three meamrements in different position were canted o~t.

Sample LDI (»m) LD2 (pm) LD3 (pm)

CSI 37 ~ 4 36 ~ 4 52 ~ 5

CGI 23~2 17~2 67~7

LSI 27~3 30~3 41~4

LGI 27~3 31~3 39~4

3. Experimental Results

The experimental results, as average values before and after the hydrogenation and post an- nealing processes are reported in Tables II to V.

It is apparent (see Tab. II) that the catalytic hydrogenation induces an increase of diffusion

length of the EFG samples which reaches values comparable with those obtained only after a 450

°

C annealing of the RF plasma pre-hydrogenated samples (see Tab. III). The EFG catalyt- ically hydrogenated material does not show any improvement after a post-annealing process.

No significant improvement of the diffusion length is instead observed in the case of EU ma- terial, both in plasma and catalytically hydrogenated samples neither after the hydrogenation

nor after the post-hydrogenation annealing (see Tabs. IV and V).

In order to understand whether this different behavior would be associated to a different density of defects which can be decorated by Pt and then made preferential sites for dissociative

chemisorption, we have measured the dislocation density by the etch pit count technique on

the EFG material before and after the catalytic hydrogenation.

Table IV. SPV meas~rements on E~rosolare lateral top samples. LDI, LD2 and LD3 are, respectwely, the average dijfwion length of the as-grown, catalytically treated and post-annealed samples. On each sample, a minim~m of three meas~rements in different positions were carried

o~t.

Sample LDI (pm) LD2 (pm) LD3 (pm)

12dA 22 ~ 2 21 ~ 2 21 ~ 2

12dB 18~2 20~2 8fil

12dC 16~2 14~2 16~2

(6)

Table V. SPV meamrements on E~rosolare lateral top samples. LDI, LD2 and LD3 are, respectively, the average dijfwion length of the as-grown, plasma treated and post-annealed samples. On each sample,

a

minim~m of three meas~rements in different positions were carried o~t.

Sample LDI (»m) LD2 (pm) LD3 (»m)

13dA 18 ~ 2 21 ~ 2 17 ~ 2

13dB 22 ~ 2 15 ~ 2 10 ~ l

13dC 19 ~ 2 17 ~ 2 11 ~ l

Fig. i SEM micrography of

an

EFG as-grown sample (magnification 200x)

The results of these measurements, carried out on SEM (Scanning Electron Microscope) pictures (see Figs. 1, 2), after purple etch removal of the massive Pt deposits, show that after the catalytic hydrogenation the apparent number of etch pits is drastically reduced. As the hydrogenation temperature is too low to allow dislocation movement, one has to conclude that

some emerging dislocation segments are no more selective etching sites, due to Pt decoration.

Apparently, these decorated dislocations are preferential hydrogen diffusion paths, with Pt which behaves as localized dissociation center of the molecular hydrogen adsorbed on it.

4. Discussion and Conclusions

The prompt increase of the diffusion length in EFG samples after catalytic hydrogenation,

without the need of a subsequent thermal annealing, supports a preliminary conclusion that the dynamics of the catalytic process are dominated by the injection of atomic hydrogen directly in correspondence with Pt-decorated emerging dislocation segments. The need of

a post-hydrogenation anneal in the plasma hydrogenated samples, moreover, indicates that in this latter case the hydrogen is ubiquitously diffused in a thin subsurface layer during the

hydrogenation process, but that it is allowed to diffuse deep into the material by the high temperature post-hydrogenation stage. In both cases, however, dislocations should behave as

preferential pathways for hydrogen to hydrogen passivable defects.

(7)

1492 JOURNAL DE PHYSIQUE III N°7

Fig. 2. SEM micrography of an EFG sample submitted to catalytic hydrogenation (magnification

200x ).

6o

so A

~

40

E ~

~ 30 '~

~

20 ,

lo Nd"10~ cm'~ Nd"10~ Nd"10~

0

o-o

Distance

from

bending

axis (cm)

Fig. 3. Diffusion length dependence

on

dislocation density before and after hydrogen passivation:

(.) Ld before hydrogenation, (A) Ld after hydrogenation.

The results of our experiments enlighten, moreover, some still undiscovered features of the hy- drogen passivation of dislocation-induced deep levels in polycrystalline silicon, which seem to fit with the multiple trapping model of Corbett et al. ill, by assuming that dislocations are unsat-

urable traps and, then, preferential paths for hydrogen diffusion and secondary hydrogenation

sources. This conclusion is well supported and implemented by the effect of plasma hydrogena-

tion on minority carrier properties in a dislocated single crystal (Fig.3), which demonstrates in fact that a substantial increase of the diffusion length Ld can be obtained by plasma hydro-

genation of dislocation-induced recombination centers.

The fact that the passivation yield, at constant process conditions, decreases with the increase of the dislocation density could not be associated to a reduced hydrogen diffusivity due to

dislocation interaction, as the hydrogen permeation increases, in fact, with the increase of the dislocation density (see Fig. 3).

Apparently, isolated dislocations are more fav(rably passivated whereas the dislocation in-

teraction process generates deep levels which can not be passivated by hydrogen.

(8)

This conclusion makes easier the interpretation of the results concerning the Eurosolare

samples (see Tabs. IV and V) which are almost insensitive to hydrogen passivation. Actually, although the EFG and EU samples were chosen with very similar compositions, they present large microstructural differences depending on the type of charge, furnace and growth cycle [5], EU samples being much less dislocated than the EFG ones.

It is then well evident that the hydrogen passivation, catalytic or plasma, has no effect on this material as the defects responsible for the electrical degradation are bulk recombination

centers (iron in supersaturation conditions and silicon nitride and iron silicide precipitates [5]) which can not be easily reached and passivated by hydrogen.

Acknowledgments

The authors would like to thank Prof. B. Pichaud of Marseille University for supplying dislo- cated samples and Dr. D. Narducci for useful discussions. This work has been supported by European Community under a Joule II Project.

References

ill Pearton S.J., Corbett S.J. and Stavola M., Hydrogen in Crystalline Semiconductors, (Springer-Verlag, 1991).

[2] Kalejs J.P., Progress in development of EFG process control in silicon ribbon for photo- voltaic applications, Material processing theory and practices, vol. 6, Series Editor, F.F.Y.

Wang (1987).

[3] Binetti S., Ratti S., Acciarri M. and Pizzini S., Study of different polycrystalline silicon materials: effect of hydrogen and deuterium passivation, Proc. 12th European Photo- voltaic Solar Energy Conference (Amsterdam 11-15 April 1994) p. 709.

[4] Pizzini S., Gasparini M. and Rustioni M., Fr. patent 82 (July 1982) p. 12588.

[5] Binetti S., Acciarri M., Savigni C., Brianza A., Pizzini S. and Musinu A., Effect of nitro- gen contamination by crucible encapsulation on polycrystalline silicon material quality,

Materials Science and Engineering B 36 (1996) 68.

[6] Mariani J.L., Pichaud B., Minari F. and Martinuzzi S., Quantitative determination of the

recombining activity of 60 °C and screw dislocations in float zone and Czochralski-grown silicon, J. Appl. Phys. 71, (1992) 1284.

[7] Acerboni S., Pizzini S., Binetti S., Acciarri M. and Pichaud B., Effect of oxygen aggre-

gation processes on the recombining activity of 60° C dislocations in Czochralski-grown

silicon, J. Appl. Phys. 76 (1994) 2703.

[8] Landolt Bornstein, New series, vol. 22 (Springer Verlag, Berlin, 1989) p. 261.

[9] Narducci D., Girardi G. and Piseri C., Preparation, Micromorphology and Stability of Tin Dioxide Thin Films, Solid State Phenomena 51-52 (1996) 435.

[10] Zundel T., Weber T., Dissociation energies of shallow-acceptor-hydrogen pairs in silicon, Phys. Rev. B 39 (1989)13549.

ill] Maffi D., Sviluppo d( un metodo di idrogenazione catalitica per la passivazione dei difetti nel silicio policristallino, Thesis Univ. Milano (1995).

[12] Ratti S., Messa a punto di una tecnica di idrogenazione in plasma per la passivazione di stati localizzati in silicio, Thesis Univ. Milano (1993).

[13] Sinha A.K., Marcus R.B., Sheng T.T. and Haszho S.E., Thermal stability of thin Ptsi

films on silicon substrates, J.Appl.Phys. 43 (1972) 3737.

Références

Documents relatifs

The present results suggest that the diffusion of copper at low temperature ( 500 os) in polysilicon is essentially a migration via structural defects which

Cette étude nous montre que la Dimedone et l’ortho-phénylènediamine deux produits commerciaux et bon marché, réagissent pour donner l’intermédiaire

2014 The analysis of the dependence of the diffusion length of minority carriers on the microstructure in polycrystalline silicon has been extended by considering

b) the discharge gap between two metallic electrodes connected in series with a localized resistance of a semiconductor electrode; c) the discharge gap between a metallic and

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

In the present paper dislocation properties were observed with the help of the dislocation luminescence spectrum discovered earlier /3>/» It should be noted, however, that

Introduction - The purpose of this paper is twofold: we shall briefly review our current understanding of the effects of atomic hydrogen on grain boundaries in silicon, and

A great emphasis on tungsten-based catalysts led to the development of a large variety of W-H 2 O 2 oxidizing systems to perform eco-friendly oxidative cleavages: from