Vacuum 71 (2003) 117–122

IBIC Studies of structural defect activity in differentpolycrystalline silicon material

V. Borjanovi!ca, M. Jak$si!cb, $Z. Pastuovi!cb, B. Pivacb,*, E. Katzc

a Faculty of Electrical Engineering and Computing, Unska 3, HR-10000 Zagreb, Croatiab Rudjer Bo$skovi!c Institute, P.O. Box 1016, HR-10000 Zagreb, Croatia

c The Jacob Balustein Institute for Desert Research, The Ben-Gurion University of the Negev, Sede Boker Campus, 84990, Israel

Received 16 June 2002; received in revised form 3 August 2002; accepted 24 September 2002

Abstract

In the research of semiconducting materials, the ion beam induced charge collection (IBIC) technique can provide

interesting and straightforward information about the different electronic device characteristics. This nuclear

microprobe technique was used for the qualitative analysis of the spatial distribution of charge collection efficiency in

several types of poly-Si material. We studied the influence of light impurities (oxygen, carbon) on electrical activity of

extended defects. It is shown that oxygen segregating close to structural defects influences their electrical activity, while

for carbon we did not observe the same effect. We demonstrated that IBIC can be applied to provide spatial

information about the position of electrically active defects and its activation during subsequent processing.

r 2003 Elsevier Science Ltd. All rights reserved.

Keywords: Silicon; Defects; Oxygen; Grain boundaries; IBIC

1. Introduction

Polycrystalline silicon (poly-Si), which can beproduced by different growth techniques, can meetboth the low-cost production and high efficiencyrequirements for solar cells. Since poly-Si wafersare inhomogeneous, various techniques have beenconsidered to track the recombination activity ofdefects across entire wafers. The electron beaminduced current (EBIC) technique in the scanningelectron microscope is a well established techniquefor the characterization of defects in semiconduct-

ing materials and devices [1]. More recently, thepossibility of making similar observations using afocused beam of high energy light ions has beendemonstrated [2–5]. As with the related and morewidely known EBIC microscopy, an energetic ionslowing down in a sample creates charge in theform of electron–hole pairs. These may drift fromtheir origin under the influence of an internalelectric field or diffuse through field-free regions ofthe sample. In the case of IBIC microscopy,however, a single MeV ion creates a dense pathof electron–hole pairs, displaces sample atoms,and may induce sample atom spallation or fission.In addition to these atomic effects, the majordifference between IBIC and EBIC are that anMeV ion will typically reach depths an order of

*Corresponding author. Tel.: +385-1-456-1068; fax: +385-

1-425-497.

E-mail address: pivac@rudjer.irb.hr (B. Pivac).

0042-207X/03/$ - see front matter r 2003 Elsevier Science Ltd. All rights reserved.

doi:10.1016/S0042-207X(02)00724-8

magnitude deeper in the sample than the electronsused in EBIC and will also undergo relativelyminimal sideways scattering. Thus deeply buriedstructures can be probed with relatively little lossof spatial resolution.

Among different possible applications of IBICthe imaging of charge-transport properties ofpolysilicon solar cells is particularly interestingwhere the properties of grain boundaries are ofconsiderable interest and have been previouslyextensively investigated by electron and optical-based probes. Donolato et al. [6] performed thefirst IBIC study on solar cells and subsequentlypresented sophisticated analytical models for theIBIC signal. Grain boundaries and other stucturaldefects are clearly visible in typical IBIC images,which are usually presented as median-energymaps. In a median-energy map, the median chargecollected from five or more ions per pixel isdisplayed as a false colour scale that effectivelyrepresents contours of charge-collection efficiency.

It was shown [7,8] that structural defects (suchas dislocations, grain boundaries and twins)together with the impurities present in solar gradematerial directly influence its electrical propertiesand therefore solar cell performance. Light non-doping impurities like oxygen and carbon are ofparticular importance because of their complexinteraction with structural defects in silicon, whichmay modify their electrical activity.

Oxygen is found in all types of silicon materialsthat have been processed in contact a with quartzcrucible. The process of oxygen incorporation intoCzochralski (CZ) single crystals has been recentlyreviewed by Borghesi et al. [9]. The float zone (FZ)silicon has very low oxygen content, typicallybelow the detection limit of IR (which is about5 � 1015 atoms cm�3), while CZ silicon containsB1018 at cm�3 of oxygen. The difference in effi-ciency of solar cells produced on CZ and FZmaterial is usually 2–3% in favour of FZ cells.This was attributed to the negative impact ofoxygen on the electronic characteristics of thematerial. On the contrary in some multicrystallinematerials, like edge-defined film-fed grown (EFG)sheets, oxygen at a certain level (up to5� 1017 cm�3) was found to be beneficial for solarcell properties. Nevertheless, the mechanism re-

sponsible for such improvement was not found. Ithas been shown that the oxygen presence passi-vates a trap level at 0.35 eV of unknown origin [10]in EFG material. In multicrystalline silicon rich invarious kinds of structural defects, oxygen tends tosegregate close to structural defects already in thecourse of material production or during thethermal treatment [11,12].

Using the deep level transient spectroscopy(DLTS) on as-received EFG material, we haveshown [13] that the concentration of electricallyactive traps, with an activation energy ofEv=0.36 eV, increases with increasing oxygencontent. We also found that temperature treat-ment might significantly change the electricalactivity of such material.

2. Experimental

Two different types of EFG Si sheets for thepresent study were grown at Mobil Solar EnergyCorp. (now ASE Americas Inc.) and at Foteks Ltd(Moscow). All of the samples were boron-doped,with resistivity 2–4O-cm and their thicknesseswere about 200 mm. ASE Americas’ EFG sampleswere grown in an argon atmosphere with orwithout addition of various amounts of CO2 gasusing a graphite die-crucible. FOTEKS’s EFGsamples were grown in vacuum using a quartzcrucible and a graphite die. The crystallinestructure of the latter crystals was demonstratedto be very similar to ASE Americas’ EFG samples[14–16].

IBIC measurements presented here were per-formed using the nuclear microprobe set-up of theRudjer Boskovic Institute in Zagreb. The nuclearmicroprobe consists of the Oxford microbeamquadrupole doublet, a data acquisition and amicrobeam control based on a SPECTOR soft-ware/hardware system. The measurements weremade using the low current irradiation geometryexplained in detail elsewhere [17]. The ion beamdiameter was about 1 mm.

All samples were cleaned and slightly etched inbuffered HF prior to contact deposition. Alumi-nium and gold were evaporated through shadowmasks on a clean surface to form ohmic (gold) and

V. Borjanovi!c et al. / Vacuum 71 (2003) 117–122118

Schottky barrier (aluminium) contacts, respec-tively. Both I–V and C–V measurements weretaken to insure the integrity of the diode char-acteristics. The DLTS spectra were taken with aSULA Technologies spectrometer. The sampletemperature was ramped from 80 to 350 K, andspectra were taken at window rates from 0.2 to50 ms.

IR measurements were performed with a Four-ier transform IR (FTIR) spectrometer (Bruker IFS113v using a Globar source) a KBr beam splitter,and a deuterated triglycerine sulfide detector.Spectra were collected from 5000 to 400 cm�1,4 cm�1 resolution, and 512 scans were averaged toimprove signal-to-noise ratio. A float zone singlecrystal silicon wafer was used as a reference toremove multiphonon absorption.

3. Results and discussion

In this study we compared IBIC images on twotypes of polycrystalline silicon materials grown bya similar method of ribbon pulling (EFG),however, one type was pulled from a graphitecontainer (EFG-A) and another from a silicacrucible (EFG-F). Fig. 1 shows the IR spectra forthe EFG material grown from the graphitecontainer. The oxygen presence in this material isclearly shown by a broad peak close to 1100 cm�1.

Moreover, IR measurements performed on bothtypes of as-received samples have always shownthe presence of oxygen in the bulk, but withdistinctly different concentrations and differentplacements in the lattice [18]. Oxygen also demon-strated a strong tendency to segregate close tostructural defects already in the phase of crystalpulling [18]. If oxygen in such a case, whendecorating structural defects, exhibits also signifi-cant influence on the electrical activity of structur-al defects, we expected the IBIC image to showdifferent charge collection efficiency (CCE) fordifferent parts of the sample, as a function ofoxygen presence and particularly close to the grainboundaries. Furthermore, IR results have shownthat, besides oxygen, these materials are particu-larly rich in carbon. It has been also shown thatoxygen and carbon, when simultaneously presentin the bulk of polycrystalline material, both tendto segregate to grain boundaries affecting there-fore the electrical activity [19]. The question alsoarises, whether carbon alone interacting withstructural defects, affects the electrical activity.

The results of IBIC measurements on an asreceived, carbon-rich and nearly oxygen free EFGsample are shown in Fig. 2. The IBIC image iscompared with the optical micrograph (taken atthe same spot), indicating that there is nomeasurable loss of CCE over the measured areaof studied sample, which includes severallarger structural defects like twin boundaries.This is in good agreement with the other findingswith EBIC, indicating that twin boundaries do notshow electrical activity. Moreover, this alsoproves that the carbon presence in the bulk, evento very high supersaturation, does not significantlyaffect the electrical activity of larger structuraldefects.

As already mentioned, samples doped withoxygen even before any thermal treatment (as-received, however, with its own thermal history)show that oxygen is not only more or lesshomogeneously distributed throughout the bulkin the form of isolated interstitial atoms, but alsodominantly segregated close to the larger structur-al defects. Furthermore, to check for a possibledistinction between decoration of dislocations inthe bulk and grain boundaries, we made diodes on

AB

SO

RB

AN

CE

(ar

b. u

nits

)

WAVENUMBERS (cm-1)

1200 1000 800 600

Fig. 1. Differential IR spectrum of EFG-Si grown from the

graphite crucible in inert atmosphere versus FZ Si oxygen and

carbon-free sample.

V. Borjanovi!c et al. / Vacuum 71 (2003) 117–122 119

the grain boundary (examining therefore the grainboundary and its vicinity) and within the grainwhich is single crystalline but with a very highdensity of dislocations (up to 107 cm�2). A cleardifference in the DLTS signal, shown in Fig. 3, ledto the conclusion that the observed oxygen-relateddefects are predominantly segregated to the largerstructural defects like grain boundaries [19]. It hasbeen already shown that, when a dislocated crystalcontains impurities in supersaturation, the pre-cipitation of impurities takes place preferentially atdislocations, influencing therefore the mechanicalproperties of Si [20].

Fig. 4(a) and (b) show IBIC images taken onsamples EFG-F grown from a quartz crucible (a)and EFG-A grown from a graphite crucible andintentionally doped with oxygen (b). As shown inthe figures most of the grain boundaries existing inthe scanned area have enhanced electrical activityi.e. a significantly reduced IBIC signal. Thereforethe first impression is that there is no significantdifference between the two samples despite theirquite different growth conditions.

Recently we reported on a comparative study ofthe impurity content in both samples grown from

graphite and quartz crucibles [18]. It has beenshown that samples grown from graphite cruciblescontain negligible amounts of interstitial oxygen,and in this case oxygen may be intentionally addedduring the growth process through an oxidizingatmosphere. On the other hand, samples grownfrom quartz crucibles contain an apparentlynegligible amount of interstitial oxygen, too.

min. max.

(a) (b)

Fig. 2. Optical micrograph (left) and 3 MeV proton IBIC image (right) of the EFG-A oxygen free sample. The area shown in the figure

is approximately 1 mm2. The white arrows in the figure show twin boundaries.

50 100 150 200 250 300 350 400

grain boundary

grain

DLT

S s

igna

l (a.

u.)

T [K]

Fig. 3. DLTS spectra taken at the grain boundary and within

the grain of EFG poly-Si sample.

V. Borjanovi!c et al. / Vacuum 71 (2003) 117–122120

However, a certain amount of oxygen is definitelypresent in the bulk but dispersed in small clustersand therefore it is IR invisible in the as-receivedsamples. Annealing of such samples at intermedi-ate temperatures produces oxygen redistribution,i.e. its precipitation. Nevertheless, the influence ofthermal treatment on C content in both types ofEFG-Si is almost negligible up to very hightemperatures. This is explained with SiI deficiencyin the Si lattice. From our unpublished DLTSresults on the effect of thermal annealing on suchmaterials we found that annealing at about 7501Csignificantly enhances the electrical activity of suchpolycrystalline material.

Fig. 5 shows the IBIC image of the EFG-Asample annealed at 7501C for two hours invacuum. As shown in the figure this temperaturetreatment produced a significant decrease in IBICresponse over all the sample area and, particularly,a widening area close to grain boundaries. Never-theless, some areas still preserve a good IBICresponse. Quite contrary to this were the findingson the EFG-F material. The figure of its IBIC

(a) (b)

min. max.

Fig. 4. 3 MeV proton IBIC image of the EFG-F oxygen-doped sample (left) and of the EFG-A oxygen-doped sample (right). Intensity

is given in relative units of a grey scale. The area shown in (a) is approximately 1 mm2, and section shown in (b) is approximately

0.5 mm2. The arrows in the figure show white areas of electrically active grain boundaries.

min. max.

Fig. 5. 3 MeV proton IBIC image of the EFG-A sample

annealed at 7501C for two hours in vacuum. Intensity is given

in relative units of a grey scale, where whiter area represents

enhanced electrical activity of defects. The area shown in the

figure is approximately 0.3 mm2.

V. Borjanovi!c et al. / Vacuum 71 (2003) 117–122 121

image after annealing at 7501C is not shown sincethe response was extremely low (barely a fewcounts scattered over the scanned area). Thisfinding for sample EFG-F is in complete accor-dance with FTIR analysis of the same material[18]. It has been shown that annealing at thisintermediate temperatures cause a significantoxygen precipitation throughout the bulk which,in turn, produces poor electronic quality of thematerial.

4. Conclusion

IBIC has been used to image the spatialdistribution of imperfections in charge collectionfor different types of EFG and EFG-like poly Si.The poly-Si materials studied were grown underdifferent growth conditions. IR measurementsshowed that oxygen present in the variousmaterials differs not only in the concentration,but also in the ways it is incorporated in the lattice(isolated interstitial position or precipitated inSiOx form and segregated close to structuraldefects). We used IBIC to study the influence ofincorporated oxygen on electrical behaviour ofextended defects present in the material. Ourfindings support the assumption that oxygen,when segregated close to structural defects, influ-ences the electrical behaviour of the ‘‘decorated’’extended defect in poly-Si material. Moreover,thermal annealing at intermediate temperaturesproduces a quite complex response of the materialthat requires a further study.

References

[1] Zoth G, Bergholz W. J Appl Phys 1990;67:6764.

[2] Donolato C, Nipoti R, Govoni D, Egeni GP, Rudello V,

Rossi P. Mater Sci Eng B 1996;42:306–10.

[3] Breese MBH, Amaku A, Wilshaw PR. Nucl Instrum

Methods B 1998;136–138:1355–60.

[4] Vittone E, Manfredotti C, Fizzotti F, Lo Giudice

A, Lorenzi A, Galassini S, Jak$si!c M. Nucl Instrum

Methods A 2002;476:607–13.

[5] Borjanovi!c V, Jak$si!c M, Pastuovi!c $Z, Pivac B, Vlahovi!c B,

Dutta J, Je$cmenica R. Sol Cells Sol En Mater 2002;72:

487–94.

[6] Donolato C, Nipoti R. J Appl Phys 1997;82:742.

[7] Acerboni S, Pizzini S, Binetti S, Accari M, Pichaud B.

J Appl Phys 1994;76:2703–10.

[8] Fedotov A, Drozdov N, Katz E, Ilyashuk Y, Mazanik A,

Ulyashin A. Phys Stat Sol A 1999;171:353–63.

[9] Borghesi A, Pivac B, Sassella A, Stella A. J Appl Phys

1995;77:4169–244.

[10] Rothagi A, Doolittle WA, Smith AW, Wald FV. Proceed-

ings of the 21st IEEE Photovoltaic Specialists Conference,

Kissmmee, USA, 1990. p. 581.

[11] Sumino K. In: Chikava J, Sumino K, Wada K, editors.

Defects and properties of semiconductors: Defect engi-

neering. Tokyo: KTK, 1987. p. 227.

[12] Pizzini S, Cagnoni P, Sandrinelli A. Appl Phys Lett

1987;51:676–8.

[13] Borjanovi!c V, Kova$cevi!c I, $Santi!c B, Pivac B. Mater Sci

Eng B 2000;71:292–6.

[14] Fedotov A, Evtody B, Fionova L, Iliashuk Ju, Katz E,

Polyak L. Phys Status Solidi A 1990;119:523.

[15] Fedotov A, Evtody B, Fionova L, Iliashuk Ju, Katz E,

Polyak L. J Cryst Growth 1990;104:186.

[16] Katz EA, Polyak LE. J Cryst Growth 1997;172:115.

[17] Jak$si!c M, Pastuovi!c $Z, Tadi!c T. Nucl Instrum Methods

B 1999;158:458–63.

[18] Pivac B, Borjanovi!c V, Kova$cevi!c I, Evtody BN, Katz E.

Sol Cells Sol En Mater 2002;72:165–71.

[19] Pizzini S, Cagnoni P, Sandrinelli A. Appl Phys Lett

1987;51:676–8.

[20] Yonenaga I, Sumino K. J Appl Phys 1996;80:734–8.

V. Borjanovi!c et al. / Vacuum 71 (2003) 117–122122