G

M

B

CIa

b

c

d

a

ARR1AA

KSSBOUE

1

satoiisqdptwtSd

te

T

0h

ARTICLE IN PRESS Model

SB-13241; No. of Pages 8

Materials Science and Engineering B xxx (2012) xxx– xxx

Contents lists available at SciVerse ScienceDirect

Materials Science and Engineering B

jou rna l h om epage: www.elsev ier .com/ locate /mseb

oron doping of silicon rich carbides: Electrical properties

. Summontea,∗, M. Caninoa, M. Allegrezzaa, M. Bellettatoa, A. Desalvoa, R. Shuklaa,b, I.P. Jainb,. Crupic, S. Militaa, L. Ortolania, L. López-Conesad, S. Estradéd, F. Peiród, B. Garridod

Consiglio Nazionale delle Ricerche – Istituto per la Microelettronica e i Microsistemi, Bologna, ItalyCentre of Non-Conventional Energy Resources, University of Rajasthan, Jaipur, IndiaConsiglio Nazionale delle Ricerche – Istituto per la Microelettronica e i Microsistemi, Catania, ItalyMIND – UB, Electronics Department, University of Barcelona, Barcelona, Spain

r t i c l e i n f o

rticle history:eceived 22 June 2012eceived in revised form1 September 2012ccepted 28 October 2012vailable online xxx

a b s t r a c t

Boron doped multilayers based on silicon carbide/silicon rich carbide, aimed at the formation of siliconnanodots for photovoltaic applications, are studied. X-ray diffraction confirms the formation of crystal-lized Si and 3C-SiC nanodomains. Fourier Transform Infrared spectroscopy indicates the occurrence ofremarkable interdiffusion between adjacent layers. However, the investigated material retains mem-ory of the initial dopant distribution. Electrical measurements suggest the presence of an unintentionaldopant impurity in the intrinsic SiC matrix. The overall volume concentration of nanodots is determinedby optical simulation and is shown not to contribute to lateral conduction. Remarkable higher room

eywords:ilicon rich carbideilicon nanodotsoron dopingptical properties

temperature dark conductivity is obtained in the multilayer that includes a boron doped well, ratherthan boron doped barrier, indicating efficient doping in the former case. Room temperature lateral darkconductivity up to 10−3 S/cm is measured on the multilayer with boron doped barrier and well. Theresult compares favorably with silicon dioxide and makes SiC encouraging for application in photovoltaic

V–vis reflection and transmittancelectrical transport

devices.

. Introduction

Silicon nanocrystals (Si-NC) embedded in a dielectric matrixhowing tunable band gap properties are presently considered asttracting top absorbers in silicon based high efficiency multijunc-ion devices [1,2]. The nanocrystals are formed by fabricating anverstoichiometric silicon compound, on which phase separationnto silicon and the wide band-gap stoichiometric compound isnduced by annealing. The wide band gap matrix has the role ofpatially separating the nanocrystals, thus giving rise to size relateduantum effects [3–5]. The multilayer approach is used to intro-uce separate control of the Si-NC size and density, and was indeedroven to be effective in the case of silicon dioxide as a dielec-ric matrix [6]. The use of silicon carbide as a dielectric matrixas introduced later, because of the lower band gap and therefore

he anticipated better conduction properties of SiC with respect toiO2 or Si3N4, more favorable in view of applications to electronicevices [2].

Please cite this article in press as: C. Summonte, et al., Mater. Sci. Eng. B (2

In a photovoltaic device, n and p doped layers are requiredo create the internal electric field that separates the photogen-rated carriers and drives them to the electrical contacts, thus

∗ Corresponding author at: CNR-IMM, via P. Gobetti 101, I-40129 Bologna, Italy.el.: +39 051 6399131; fax: +39 051 6399216.

E-mail address: caterina.summonte@cnr.it (C. Summonte).

921-5107/$ – see front matter © 2012 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.mseb.2012.10.030

© 2012 Elsevier B.V. All rights reserved.

preventing recombination. If Si-NC are used as absorber material inthe device, the bandgap of doped layers should be large enough, notto squander the energy of carrier photogenerated in the nanodots.Doped Si nanodots will in principle meet the requirement, and aproof of concept on the topic has been reported for the case of SiO2matrix [7,8]. For Si-NC in SiC matrix, the B doped Si-NC:SiC materialhas shown to form a heterojunction with n-type c-Si [9], althoughit is hard to distinguish whether the acceptor level involved in theobserved built-in voltage formation is located within the Si-NC, theSiC matrix, or at the interface.

If doping is introduced in the Si-NC based absorber, this will alterthe carrier collection properties, through modification of the elec-tric field across the device; introduction, or possibly passivation,of electronic defects; modification of the resistive losses and soon. Prediction of such effects implies information on the electronicproperties of the composite (Si-NC and dielectric matrix) material.Effective B doping of Si-NC was already observed [10,11], althoughin SiO2 matrix it suffers from segregation at the Si/SiO2 interfaceand in the bulk of the Si-NC, with electrically active configurationonly in the latter case [12]. The role of boron was also demonstratedto depend on the nanodot size [13]. Less attention has been paidto the electrical properties of the matrix, and to its contribution

012), http://dx.doi.org/10.1016/j.mseb.2012.10.030

to the overall electrical performance. In the case of Si-NC in SiO2,the insulating matrix allows assuming that transport takes placethrough silicon related subdomains – in fact, it becomes detectableonly for sufficiently close Si-NC spacing [14]. In the case of the

ING Model

M

2 ce and

Ssttaaflo

NiitsStenc

2

awt1tiRfmtB(fae1Rwiawis3tnolstne(

(secpor

ARTICLESB-13241; No. of Pages 8

C. Summonte et al. / Materials Scien

i-NC:SiC system, due to the lower band gap and less marked phaseeparation, a parallel of conducting paths is to be expected, andhe different contributions are less evident to separate. This fur-her complicates the analysis of doped materials. The situation islso affected by possible diffusion of the dopant within the matrixnd among sublayers during thermal treatments [9]. In summary,urther investigation seems highly significant due to the limitediterature existing in the case of SiC matrix [9,15], particularly ifbtained by PECVD [16–19].

In this paper, we investigate the electrical behavior of the Si-C:SiC system with B doping, where B doping is introduced either

n the SRC or in the SiC sublayers, or both (modulated doping),n order to shed light on the physics of the system, and identifyhe most favorable doping scheme. The comparison with referenceingle layers will allow to evidence the effect of the presence ofi-NC embedded in the matrix, and the role of interfacial states;he electrical results will be discussed also basing on the differ-nt crystallization occurring in a massive SRC [16] compared to theanosized sublayer, as a consequence of the fact that small siliconlusters tend to remain amorphous [20,21].

. Experimental procedure

The Si nanodot fabrication was based on the multilayerpproach. Silicon rich carbides (SRC)/stoichiometric SiC multilayersere fabricated by Plasma Enhanced Chemical Vapour Deposi-

ion (PECVD), on quartz, and bare and oxidized Si substrates, at3.56 MHz, 24 mW/cm2, 0.95 mbar pressure, and 160 ◦C substrateemperature Ts, using SiH4, CH4, and H2 as gas precursors. The nom-nal Si atomic fraction x in the SRC was 0.65, obtained with gas ratiog = SiH4:CH4:H2 = 6.1:70:10 (flow rates in sccm). Rg was 2:70:10

or SiC. Boron doping was achieved by introducing B2H6 in the gasixture, 0.8% and 1%, in SRC and SiC respectively, with respect

o SiH4. Three multilayers, M1, M2, M3, were fabricated, in whichoron was introduced either within the SRC (M1) or within the SiCM2) or in both layers (M3). An intrinsic multilayer (M0) was alsoabricated, using 1.0 mbar pressure, and Ts = 350 ◦C, Rg = 11.4:70:10nd 2.6:70:10 for SRC (x = 0.75) and SiC respectively. The multilay-rs were fabricated by starting with a 20 nm SiC layer that shrinks to0 nm after annealing in correlation with hydrogen evolution, seeef. [22] for details. Then we deposited 30 SRC + SiC double layersith thicknesses regulated in order to obtain 3 nm after anneal-

ng for both, by taking into account the specific shrinkage of SiCnd SRC upon annealing. An a-Si:H sacrificial layer, 20 nm thick,as then deposited on the top of the samples. The reason for this

s given below. After multilayer deposition and prior to the a-Si:Hacrificial layer deposition, the doped samples underwent an in situ50 ◦C, 4 h annealing, to avoid sample blistering during subsequenthermal treatments, mainly because the available annealing fur-ace can be controlled only above that temperature. Test samplesf SiC and SRC, either doped or undoped, about 200 nm thick singleayers samples, and undoped multilayers, were also fabricated onimilar substrates with the same procedure. The optical gap E04 forhe annealed SiC material is 2.3 eV. The Tauc plot procedure couldot be applied, due to the strong non-linearity of the Tauc plot. Ref-rence a-Si and a-SiC layers were obtained after dehydrogenation600 ◦C, 4 h) of the corresponding PECVD deposited material.

Subsequently, all samples underwent a two step annealing600 ◦C, 4 h + 1100 ◦C, 30′ plus ramps) in fluent N2:O2 = 10:1 atmo-phere. The low temperature step is required to produce hydrogenvolution, while the high temperature step results in SiC and Si

Please cite this article in press as: C. Summonte, et al., Mater. Sci. Eng. B (2

rystallization. Due to the presence of oxygen during the annealingrocess the silicon sacrificial layer partially oxidizes. This surfacexide is then removed in HF, and the residual unoxidized Si layer isemoved in a 2% TMAH solution at 40 ◦C. This procedure is needed to

PRESS Engineering B xxx (2012) xxx– xxx

avoid formation of a surface SiCyOz compound, with stoichiometrydepending on the composition of the underlying material, that con-sumes part of the initial double-layers, and that cannot be removedwith standard microelectronics wet etching (either HF, TMAH, orCP4). Details on the identification and elimination of the SiCyOz

compound can be found in [23], where the effect of such unwantedcapping layer on electrical measurements is also illustrated.

The lateral conductance as a function of temperature was mea-sured using parallel plane Ni evaporated contacts spaced 1 mm. Atwo terminal configuration was used in a K20A vacuum cryostatby MMR Technologies. The current was measured using a Keithley219 amperometer. The ohmicity of the contacts was checked in therange −30 V/+30 V.

The reflectance and transmittance (R&T) spectra were measuredusing an Avantes UV–vis fiber optics spectrophotometer and aHP8542A diode array spectrophotometer. The spectral data wereelaborated by means of the computer code Optical that treats themultilayer structure using the Jones matrix – Fresnel coefficientapproach, and allows the introduction of the Bruggeman effectivemedium approximation (EMA) model in the description of the spec-tral refractive index of the layers. The materials used in the EMAmixture were amorphous and microcrystalline silicon, with n-kspectra taken from Refs. [24,25] respectively, and nominally micro-crystalline SiC [22]. In this last case, the n-k spectra were deducedfrom R&T measurement taken on a single SiC sample deposited onquartz and annealed at 1100 ◦C as described above, and fitted usingthe modified Tauc–Lorentz model proposed by [26] further mod-ified to finely account for defect excess absorption [Allegrezza, tobe published]. The simulation also revealed the presence of a lowdensity surface layer, that was taken into account while obtainingthe n-k spectra of �c-SiC.

Infrared transmittance was measured using a Thermo Nicolet5700 Fourier Transform Infrared spectrometer (FTIR). The IR trans-mittance data were simulated using Optical (see above), in whichthe spectral n-k of silicon was simulated using a combination of17 Lorentz oscillators, to account for the vibrational spectrum of c-Si. The doped surface region of the wafer was simulated using thesame set of oscillators with the introduction of a free carrier absorp-tion Drude component. A measurement of the Drude component inthe Si wafer substrate was achieved by removing the deposited Si-NC layers via dry etching in an SF6:O2 plasma with 64:22 (flowsin sccm) gas flow ratio, at 75 mTorr, 80 W. The etch rate showeda marked dependency on composition and was determined to be23 nm/min for stoichiometric SiC, 40 nm/min for Si0.65C0.35, andaround 700 nm/min for c-Si [27]. The etching duration was chosenin order to uncover the c-Si substrate surface.

XRD patterns were collected in out-of plane grazing incidencegeometry, where the incident angle, i.e. the angle between theincoming beam and the sample surface, was kept constant and at asmall value (0.8◦) while the detector moved (2� scan) in the planenormal to the surface The measurements were performed using theSmartLab diffractometer (Rigaku), equipped with a rotating anode(Cu K˛ � = 1.54180 A) followed by a parabolic mirror, to collimatethe incident beam and Soller slits (divergence of 0.114◦) in front ofthe detector to collimate the diffracted beam.

High Resolution Transmission Electrom Microscopy (HRTEM)was performed by means of cross-section energy-filtered TEM(EFTEM) images, taken using a JEOL 2010 instrument, operating at200 keV FEG, equipped with a Gatan image filter (0.8 eV resolution).

3. Results and discussion

012), http://dx.doi.org/10.1016/j.mseb.2012.10.030

3.1. Structural results

Si-NC formation was observed by TEM. Fig. 1(a) shows theHRTEM image of a multilayer with intrinsic SiC barrier and SRC

ARTICLE IN PRESSG Model

MSB-13241; No. of Pages 8

C. Summonte et al. / Materials Science and Engineering B xxx (2012) xxx– xxx 3

F REM( spatiai to the

wwcswpihtriiio4

ddtdssisMa

Fwo�a

ig. 1. (a) High resolution TEM image of the intrinsic multilayer. (b) FFT of the Hgreen) were highlighted. (c) Results of applying the filter to the FFT, showing the

nterpretation of the references to color in this figure legend, the reader is referred

ell after annealing. The image shows a polycrystalline network,here the multilayer network is not maintained. In fact, unlike the

ase of Si-NC in SiO2 matrix, the multilayer structure is known tomear out after annealing for the case of SiC matrix [9]. In Fig. 1(b)e reported the Fast-Fourier-Transform (FFT) of Fig. 1(a), whereoly-crystalline ring patterns from Si (1 1 1), corresponding to the

nterplanar distance of 0.314 nm, and 3C-SiC (1 1 1), 0.251 nm, wereighlighted, respectively, by red and green circles. A filter is appliedo the FFT, to select only Si (1 1 1) and 3C-SiC (1 1 1) reflections, thusevealing the position of the crystals originating those reflectionsn the HREM image. The results of the filter are shown in Fig. 1(c)n false colors. The regions with Si and SiC lattice fringes are shownn red and in green respectively. The figure confirms the presencef nano-crystals of both materials, with nanodot size of the order–6 nm.

To quantitatively follow the nanodot formation upon intro-uction of boron, the intrinsic and doped annealed multilayerseposited on c-Si substrate were analyzed using X-ray diffractionechnique [28,29]. The results are reported in Fig. 2. The presence ofiffracted peaks shows evidence of Si and 3C-SiC nanocrystals in allamples. The peaks are enhanced compared to the intrinsic case, inpite of the slightly lower silicon content (0.65 instead 0.75). A sim-lar occurrence was also detected on B doped compared to intrinsic

Please cite this article in press as: C. Summonte, et al., Mater. Sci. Eng. B (2

ingle layers [30]. Very similar spectra are obtained for samples2 and M3, whereas in the case of M1, lower volume for both Si

nd SiC materials is observed. As the peak intensity also depends

20 30 40 50 60 70 80

0.5

1.0

1.5

Si SiC

6.8 5.06.5 4.86.1 4.4

M0 M1 M2 M3

Si(311 )

Si(331 )SiC(220 )SiC(200 ) SiC(311 )

Si(220 )

SiC(111 )

XRD

inte

nsity

(cps

x 1

000)

2 (de g)

Si(111 )

ig. 2. X-ray diffraction of multilayers deposited on c-Si. Curves M1, M2, and M3ere upshifted by 50, 400 and 200 cps respectively for clarity. The angular position

f relevant Si and �-SiC diffraction peaks, and the nanodot size (in nm) of Si and-SiC for doped multilayers,as deduced from the FWHM of (1 1 1) diffraction peaks,re indicated.

image where rings corresponding to silicon reflections (red) and silicon carbidel distribution of Si (red) and 3C-SiC (green) nano-crystals in the HREM image. (For

web version of the article.)

on thickness, this point will be further commented below (Section3.2).

Using the Scherrer expression [31] on the basis of the fullwidth at half maximum intensity of the (1 1 1) diffraction peak,the calculated Si and SiC nanodot size is 6.4 ± 0.4 and 4.7 ± 0.3 nmrespectively. These values are accurate within 4–5%, as may beshown accounting for the uncertainties in the full widths at halfmaximum intensity. It is important to note that the Scherrer equa-tion is itself an approximation, since the crystals within a givensample may have a size distribution. Therefore, one can speak onlyof an average crystal size within the sample and the presence ofsmaller nanodots cannot be completely excluded. The results allowto conclude that effective crystallization is achieved in presence ofB doping for samples fabricated on c-Si substrate.

3.2. Optical measurements

The silicon crystallized fraction was also checked by Reflectanceand Transmittance (R&T) spectroscopy. Fig. 3 reports the measuredand simulated R&T spectra for the three multilayers on quartz sub-strate, and the R spectra on c-Si substrate. Several features can beobserved. In the UV spectral range, due to high light absorption ofthe materials in this region, the absorption length is less than thesample thickness. The experimental R spectrum then probes onlythe topmost region, and is independent of substrate. Fig. 3 showsthat actually similar R is observed for all samples in this region,the differences being due to the details in composition and rough-ness of the residual surface oxide [23]. For longer wavelengths, theabsorption decreases, and, at a wavelength that depends on absorp-tion and thickness, the incident light reaches the substrate and ispartly reflected. This gives rise to the onset of transmittance T in thecase of a transparent substrate such as quartz, and of an interferencepattern on R&T. Such wavelength is about 350 nm for all samplesin Fig. 3, showing a substantial similarity between the samples. Theinterference period is related to thickness and refractive index ofthe thin film. In Fig. 3, slightly blue shifted fringes are observed forsample M1, indicating a slightly lower thickness, which is in agree-ment with the lower diffracted XRD intensity observed in Fig. 2.The contrast of the interference pattern depends on the differencein refractive index between layers and substrate, and is damped byabsorption. For this reason, as the average refractive index of SRCbased multilayer is more similar to c-Si than to quartz, the con-trast observed on c-Si substrate is lower than on quartz. Sample M1also shows a moderately lower contrast, with respect to the other

012), http://dx.doi.org/10.1016/j.mseb.2012.10.030

samples, more evident on c-Si, which is indicative of presence of ahigher Si concentration, with a higher absorbing amorphous siliconcomponent, again in agreement with the lower SiC and Si diffractingvolumes obtained with the XRD analysis.

ARTICLE IN PRESSG Model

MSB-13241; No. of Pages 8

4 C. Summonte et al. / Materials Science and Engineering B xxx (2012) xxx– xxx

Table 1Structure used to simulate the R&T spectra measured on quartz and the R spectra taken on c-Si reported in Fig. 3. The data for the intrinsic multilayer (=M0), not includedin Fig. 3, are also reported. The structure used in the simulation was: (1) surface layer; (2) bulk of the sample, (3) 10 nm SiC buffer layer (fixed in the simulation) and (4)substrate (either quartz or c-Si). The variables were thickness (d) and compositions of layers (1) and (2). The composition of the Si-NC layer (2) was obtained using an EMAmixture of �c-Si [25] a-Si [24], and microcrystallized SiC. XC is the crystalline silicon fraction with respect to the total Si content in the bulk of the samples. The errors barsare given and discussed in the text.

On quartz substrate On c-Si substrate

Surface layer Bulk of the sample Surface layer Bulk of the sample

d (nm) SiC (%) a-Si (%) d (nm) SiC(%)

a-Si(%)

�c-Si(%)

XC d (nm) SiC(%)

SiO2

(%)a-Si(%)

�c-Si(%)

d (nm) SiC(%)

a-Si(%)

�c-Si(%)

XC

M0 5.9 93 7 148.5 69 23 8 0.26 6.3 70 20 10 – 170.0 67 22 11 0.33.1

.8

.2

sc((Taii

Fm

M1 4.5 100 0 126.6 71 22 7 0.24 7M2 4.5 100 0 148.1 74 17 9 0.35 7M3 5.7 100 0 150.4 75 19 6 0.24 6

Such qualitative features were quantified by means of opticalimulation, that were performed using the computer code Opti-al [32]. All spectra in Fig. 3 were simulated as a stack of 4 layers:1) surface layer; (2) bulk of the sample, (3) 10 nm SiC buffer layerfixed in the simulation) and (4) substrate (either quartz or c-Si).

Please cite this article in press as: C. Summonte, et al., Mater. Sci. Eng. B (2

he variables were thickness (d) and composition of layers (1)nd (2). The composition of the Si-NC layer (2) that is the mostnfluent parameter on the shape of the interference patterns shownn Fig. 3, was obtained using an EMA mixture of �c-Si [25], a-Si [24],

200 400 600 800 10000

20

40

60

80

100

R on sili con

M2 measured computed measured computed

R on quartz

T on quartz

R,T

Wavelength (nm)

200 400 600 800 10000

20

40

60

80

100

R on quartz

R on silicon

T on quartzM1 measured computed measured computed

R,T

Wavelength (nm)

200 400 600 800 10000

20

40

60

80

100

R on sili con

M3

R on quartz

T on quartz

measured computed measured computed

R,T

Wavelength (nm)

ig. 3. Measured (symbols) and computed (lines) spectra measured on the annealedultilayers: R&T on quartz, and R on c-Si.

45 45 – 10 136.1 53 24 23 0.4950 50 – – 181.9 74 6 20 0.77

100 0 – – 176.5 70 10 20 0.67

and microcrystallized SiC. XC is the crystalline silicon fraction withrespect to the total Si content in the bulk of the samples. The erroron the material fractions in the bulk of the samples is in the range±1–5% (absolute values), higher for higher fractions. The surfacelayer (1) is necessary to obtain a good fit, because all the R spectrashown in Fig. 3 exhibit lower UV values than the ones calculatedby considering the composition obtained for layer (2). This indi-cates the presence of a residual low density surface layer, whichis interpreted as a SiC-rich surface layer resulting from the TMAHetch process that was used to remove the a-SiH sacrificial layer[23]. Much larger errors exist on the determination of the compo-sition of the surface layer since, due to its very low thickness, it doesnot produce an interference pattern by contrast to layer (2). Basi-cally, the R&T spectra give a precise indication of the presence ofa low-density surface layer, however, the EMA mixture needed toobtain such layer is not univocally defined. The error on total mul-tilayer thickness is ±4 nm over the total sample thickness, mainlydue to uncertainties on the surface layer composition. The samplestructure obtained is reported in Table 1.

From the table, it is seen that silicon crystallization is enhancedon c-Si substrates (columns XC). The reason of such result is notclear. Substrate driven epitaxial regrowth is discarded following aprevious experiment (not reported) that indicated that a 20 nm SiCbuffer layer, as in the present case, is a safe thickness to preventthis occurrence. The data also indicate the presence of a residualSi O component in the surface layers in the case of c-Si substrates,as also separately detected by XPS [23].

In order to compare the optical and structural results, the XRDdetected Si (1 1 1) peak area measured on samples on c-Si substrates

012), http://dx.doi.org/10.1016/j.mseb.2012.10.030

and normalized by sample thickness, and the crystallized fractionsobtained by R&T, both on c-Si and on quartz substrates (see columnsXC in Table 1), are plotted in Fig. 4. Similar trends are obtainedoptically and by XRD for the c-Si substrate case, supporting the

M0 M1 M2 M30

1

2

3

4 XRD on c-Si

XR

D S

i(111

) pe

ak a

rea,

a.u

(nor

mal

ized

)

SAMPLE

0.0

0.2

0.4

0.6

0.8

1.0

Opt

ical

c-S

i Fra

ctio

n Optical, on c-Si Optical, on quartz

Fig. 4. Silicon crystallized fraction obtained by optical simulation, vs the Si (1 1 1)peak area obtained by XRD and normalized by sample thickness.

IN PRESSG Model

M

ce and Engineering B xxx (2012) xxx– xxx 5

vcs

3

Sadl

V

wcmiaiivnorimbvbfpmaltoqHtiib

3

ogSepdsoitmrurmfBtO

4000 3500 3000 2500 2000 1500 1000 5000

10

20

30

40

50

60

ND=1.6E19 cm-3

ND=2.5E19 cm-3

ND=4.2E19 cm-3

measu red simulated

M2

M1

M3

c-Si, 1 cm

Tran

smitt

ance

(%)

Wavenumber (cm-1 )

c-Si

cSi + Drude

Fig. 5. Mid-infrared transmittance of silicon substrates after multilayer fabricationand removal. Untreated reference substrate is reported for comparison. The sketch

ARTICLESB-13241; No. of Pages 8

C. Summonte et al. / Materials Scien

alidity of the optical analysis. In turn, optical data show a lowerrystallized fraction on quartz substrates, indicating that a differenttructure is obtained in this case.

.3. Numerical evaluation of the volume fractions

To further check the validity of optical analysis, the intendedi/SiC composition was separately determined by computing the Sind SiC volume, basing on atomic composition, atomic mass, andensity of the two materials. The Si/SiC nominal fraction in the well

ayer is given by:

i = Ei(m0ri + mi)�i

· 1∑

i(Ei(m0ri + mi)/�i)(1)

here i = 0 refers to silicon and i = 1 refer to SiC; Ei is the atomiconcentration of the element (Si, C), �i and mi are the density andolecular masses of Si and SiC, r0 is zero and r1 is the Si-to-C ratio

n SiC (=1). In the case of SRC of composition SixC1−x, then E0 = xnd E1 = 1 − x. Eq. (1) is generally valid for the case of an oversto-chiometric compound that undergoes complete phase separationnto the excess element and the stoichiometric compound [19]. Theolumes obtained using Eq. (1) are further modified if the compo-ents suffer different shrinkage upon annealing. In fact, we havebserved that SiC suffers from remarkable (about 50%) volumeeduction upon 1100 ◦C, 30′ annealing, whereas only a 5% shrinkages detected for Si [19,22]. Finally, the volume fractions in the overall

ultilayer must take into account the relative thickness (dB, dW) ofarrier and well. Using such formulation, the nominal Si and SiColume fractions in the annealed multilayers were determined toe 28% and 72%. This last number shows a good match with the SiCraction determined by optical simulation, see Table 1, thus sup-orting the single layer approach to simulate the 30 + 30 annealedultilayer structure, and the overall validity of optical analysis. We

lso obtained that the Si-to-SiC volume ratio in the annealed wellayer is 58:42. By combining these numbers with the silicon crys-allized fraction (Table 1), it is obtained that the crystallized siliconccupies about 14–20% of the well layer for the case of samples onuartz substrates, which were used for electrical measurements.owever, this evaluation is to be considered as a rough estima-

ion, as it is affected by the error in the crystalline silicon fractionn the EMA, and because it assumes sharp barrier/well interfacesn annealed samples, which is a feature that for Si/SiC has not yeteen confirmed.

.4. Interlayer diffusion: FTIR measurements

A concern is represented by the possibility that diffusion of Bccurs between adjacent layers during the annealing, along therain boundaries of SiC rather than in the bulk of the crystallizedi and SiC domains. To the best of our knowledge, little literaturexists on the diffusivity of B in microcrystalline SiC in the tem-erature range used to produce Si-NC. Very limited interlayer Biffusion was previously observed by SIMS in SRC deposited byputtering for the same temperature and a shorter (9 min insteadf 30 min) annealing time than in the present case [9]. To gainnsight on this aspect we analyzed the infrared transmission ofhe silicon substrates that were used for multilayer fabrication,

easured after multilayer remotion by dry etching. The results areeported in Fig. 5. In the figure, the transmittance spectrum of anntreated substrate is reported for comparison. The figure shows aemarkable decrease of transmission for the substrates used for theultilayer fabrication, which can only be simulated by introducing

Please cite this article in press as: C. Summonte, et al., Mater. Sci. Eng. B (2

ree carrier absorption in the substrate. This indicates diffusion of through the multilayer. Based on these data, the boron concen-ration in the substrate was estimated using the computer codeptical [32]; see also Section 2. The transmittance of the untreated

represents the structure used in the simulation. Drude parameters used in the simu-lation are indicated in the figure. The corresponding B concentrations are 1.6 × 1019,2.5 × 1019, 4.2 × 1019 cm−3.

substrate could be simulated using the analytical refractive indexfor silicon, obtained as a sum of 17 Lorentz oscillators, each relatedto a specific vibrational frequency of c-Si. To simulate the treatedsubstrates, a surface layer was used, which was modeled usingagain the analytical refractive index of silicon, yet adding a freecarrier Drude component. The simulations were computed using abox profile obtained by fixing the thickness to 600 nm, obtainedfrom literature data for B diffusion in Si [33] for the annealingconditions used in this paper, and adjusting the Drude parame-ters. Although the approximation is very rough, an estimate of theamount of dopant in the substrate is obtained. In particular, totalconcentrations from 2 × 1019 to 4 × 1019 cm−3 were obtained, thatare approximately one order of magnitude lower than the nominalconcentration incorporated in the film. In particular, a comparisonof the M1 with the M2 substrate, that incorporate the same doseand differ basically by the 10 nm SiC buffer layer, that is doped onlyfor M2 sample, shows that such intrinsic 10 nm SiC buffer actuallylimits the B diffusion to a factor 0.6, while the higher dose detectedfor the M3 sample indicates that the process in the other two casesis source-limited. Further investigation is needed, to obtain a com-plete picture of the diffusion behavior, however, from these datawe can conclude that remarkable, grain boundary driven, diffusionactually occurs between the layers; however, a flat profile all overthe sample is not obtained, and the samples retain memory of theinitial dopant distribution. The result is in agreement with SIMSobservation on fairly similar samples [9].

3.5. Electrical measurements

Electrical conductivity was measured as a function of tempera-ture for all samples. I–V measurements taken at room temperaturefrom −30 to +30 V, reported in Fig. 6, showed an ohmic behav-ior in almost all samples, except in two cases for which the verylow conductivity was close to the detection limit of our apparatus(amorphous SiC, and multilayer M2). This result indicates that at theinvestigated temperature the conduction mechanism is not barrierlimited, and we are observing continuous conduction paths alongthe material. This suggests that, in almost all cases, the conduc-tivity of crystallized SiC is sufficient to guarantee transport alongthe dielectric matrix, and that silicon nanodots, which, as seen byUV–vis spectroscopy, occupy a volume of about 14% within the well

012), http://dx.doi.org/10.1016/j.mseb.2012.10.030

layer, primarily act as carrier scattering centers.The Arrhenius plots (conductivity as a function of inverse tem-

perature) of the samples are reported in Fig. 6, together with theplots of reference samples (single layer intrinsic amorphous and

ARTICLE IN PRESSG Model

MSB-13241; No. of Pages 8

6 C. Summonte et al. / Materials Science and Engineering B xxx (2012) xxx– xxx

Table 2Conductivity activation energy Ea and preexponential factor �0 obtained from the Arrhenius plots reported in Fig. 6: a-Si, dehydrogenated amorphous silicon; a-SiC, dehydro-genated amorphous silicon carbide; i-B (=intrinsic barrier), microcrystallized SiC; B:B (=B-doped barrier), boron doped microcrystallized SiC; i-�cSi, intrinsic microcrystallizedsilicon; i-W (=intrinsic well), microcrystallized SRC; W:B (=B-doped well), boron doped microcrystallized SRC; M0, intrinsic multilayer; M2, multilayer with boron dopedbarrier; M1, multilayer with boron doped well; M3, multilayer with boron doped barrier and well.

a-Si a-SiC i-B B:B i-�c-Si i-W W:B M0 M1 M2 M3

−4

−4

.4

mSaob

�

wsf

btttdrsoos

welcedditc

Fiasbsblm

�0 (S/cm) 9 × 10 14 0.015 29 93Ea (eV) 0.16 0.4 0.19 0.32 0

icrocrystallized SiC, SRC, and Si; single layer B-doped SiC andRC; intrinsic multilayer). In the investigated temperature range,n activated behavior with a single exponential dependence isbserved. The temperature dependence of conductivity � can thene described by the equation:

= �0 · e−(Ea/kT) (2)

here T is the absolute temperature and k is the Boltzmann con-tant. The conductivity activation energy Ea and preexponentialactor �0 for all measured samples are reported in Table 2.

If B is introduced into the single layer of SRC (comparisonetween i-W and W:B), we observed an increase of the conduc-ivity prefactor from 0.8 to 3 S/cm and a decrease of Ea from 0.27o 0.15 eV, that results in an increase of room temperature conduc-ivity by three orders of magnitudes (see Fig. 6), indicating efficientoping of the material. In this case, the silicon volume in the mate-ial equals 57%, with a crystallized fraction, determined by R&Tpectroscopy, of 0.65. As these data indicate the presence of anverall 37% crystallized volume in the material, we are probablybserving conduction along a continuous doped microcrystallizedilicon path.

Upon B doping, the SiC shows an increase of activation energy,hich is the opposite of what one would expect upon doping. How-

ver, the nominally intrinsic SiC (=i-B in Table 2) shows a veryow activation energy (0.19 eV), and a 10−5 S/cm room temperatureonductivity, that is rather high for an intrinsic material with annergy gap of 2.4 eV. Such behavior could be ascribed to either con-uction through a network of defects, or the presence of unintended

Please cite this article in press as: C. Summonte, et al., Mater. Sci. Eng. B (2

oping impurity, possibly oxygen or nitrogen. A similar effect wasndeed reported for microcrystalline silicon as due to oxygen con-amination [34], while nitrogen is known to act as unwanted n-typeontaminant in monocrystalline SiC.

-40 -20 0 20 40

-0.7

0.0

0.7

Nor

mal

ized

Cur

rent

(ar

b.u.

)

Voltage (V)

ig. 6. I–V measured on all samples, differently normalized in order to arrange themn the same plot. The inset shows the enlargement measured at low voltage bias. (�)-Si, dehydrogenated amorphous silicon; ( ) a-SiC, dehydrogenated amorphousilicon carbide; (�) i-B (=intrinsic barrier), microcrystallized SiC; (�) B:B (=B-dopedarrier), boron doped microcrystallized SiC; (♦) i-�cSi, intrinsic microcrystallizedilicon; (�) i-W, microcrystallized SRC (=intrinsic well); (�) W:B (=B-doped well),oron doped microcrystallized SRC; (�) M0, intrinsic multilayer; ( ) M2, multi-

ayer with boron doped barrier; (�) M1, multilayer with boron doped well; (�) M3,ultilayer with boron doped barrier and well.

0.8 3 0.07 0.04 2.4 × 10 7.20.27 0.15 0.2 0.28 0.33 0.25

In the intrinsic multilayer, the silicon subfraction is 31%, 26% ofwhich is crystallized, for a total crystallized silicon volume fractionof 8% (see line 1 of Table 1, samples on quartz). Conduction alongthe crystallized silicon subcomponent is therefore prevented, andwe are probably observing the continuous conduction path throughthe SiC matrix (=i-B in the Table 2); in fact, the electrical behaviorof the overall material is rather similar, see Fig. 6.

If boron doping of the barrier of the multilayer is performed(multilayer M2 in the table), we observed an unexpected drop of theprefactor by more than two orders of magnitude (from 7 × 10−2 to2 × 10−4 S/cm). The room temperature conductivity of the materialdecreases by three orders of magnitude, see Fig. 6. This occurrencecan be justified if we assume that the introduction of boron actsin the sense of compensating the donor level, but not to a levelhigh enough to produce p-type doping as in the SiC alone. In thecase of oxygen contamination in microcrystalline silicon, low levelB doping was indeed found to act as compensator [35]. The B con-centration is probably lower than in the single SiC layer doping,because of diffusion in the adjacent well layers. However, even sup-posing homogeneous redistribution of B, that we have seen it is notthe case, this would imply a drop of B concentration by a factor of 2,which does not justify the observed drop in conductivity. Moreover,in the multilayer the defect concentration is much higher, the defectbeing located mainly at the grain boundaries of SiC and between Siand SiC grains [36]. The drop in conductivity can be interpreted ifwe assume that B preferentially locates at the additional Si-relateddefects, with the formation of non-electrically active B-defect com-plexes, thus leaving not enough source for B doping of the SiC. Anhypothesis is also that in this particular case a continuous tissueof amorphous SiC dominates the electrical properties. In fact, sin-gle layer intrinsic dehydrogenated amorphous SiC shows electricalproperties not very different from the case of the M2 multilayer(see Table 2 and Fig. 7). In this scenario, where a possible dopinglevel due to boron is shorted by the unsaturated dangling bondconcentration, would imply justifying the different crystallizationbehavior on c-Si and on quartz discussed above, and the differ-ent crystallization and doping properties that are observed in theother B doped multilayers. We note that by R&T analysis we werenot able to detect an amorphous SiC component, partly becausewe do not have a suitable standard. However, the simulation couldbe successfully performed using the SiC n-k spectral data obtainedon 1100 ◦C crystallized sample, see Section 2 for details, and thatthe introduction in the mixture of the n-k data for ion implantedamorphized SiC would not improve the simulation. Finally, we notethat oxygen was already shown to hinder SiC crystallization, withfavorable effects on solar cell performance [37].

In the case of the multilayer with doped well (=multilayer M1in the table), the overall boron content is similar to the case of thedoped barrier. However, the conductivity is three orders of magni-tude larger. This confirms that, in spite of the remarkable diffusionoccurring, the material retains memory of the initial distributionof boron within the well layer. The different result is not related

012), http://dx.doi.org/10.1016/j.mseb.2012.10.030

to the increased crystallization – in fact, optical results indicatethat the Si crystallized fraction on quartz is similar, about 25% inboth cases (see Table 1, col XC, and Fig. 4). The result is thereforeattributed to efficient boron doping of the well layers, in either of

ARTICLE IN PRESSG Model

MSB-13241; No. of Pages 8

C. Summonte et al. / Materials Science and Engineering B xxx (2012) xxx– xxx 7

2.6 2.8 3.0 3.210-10

10-9

10-8

10-7

10-6

1x10-5

1x10-4

10-3

10-2

10-1

2.6 2.8 3.0 3.22.6 2.8 3.0 3.2

a-SiC

a-Si

Intrins ic materials

d(S/c

m)

1000/T (K-1)

Doped multi lay ers

1000/T (K-1)

M3

M1

W:BM0

i-W

µc-SiC = i -B

M2

Doped single laye rs

µc- Si

B:B

1000/T (K-1)

a b c

F sic (ca layer.

iSitr

TbarcaSdpbno

4

b1tbdRstaaogtimpddahba

[

[

[

[

[[[

[

[

[

ig. 7. Arrhenius plots of multilayers and reference samples. (a) Single layer intrinnd SRC. (c) Doped multilayers. Each curve is labeled with the name of the specific

ts components: the phase separated SiC matrix, the crystallizedi domain, or the residual amorphous silicon network surround-ng the nanodots. An important role of such amorphous region tohe conduction properties of the sample was indeed previouslyeported for a somewhat different hydrogenated material [38].

When both barrier and well are doped (=multilayer M3 inable 2), further increase in the room temperature conductivityy three orders of magnitude was observed. 7.2 S/cm and 0.25 eVre measured as preexponential factor and activation energyespectively, that are compatible with efficient B doping of micro-rystallized SiC. We can conclude that the B supply is sufficient toct both as defect passivator and as dopant of both the amorphousi network and the SiC matrix, and makes the material suitable forevice application. Preliminary results on photoluminescence ofassivated samples seem to confirm the proposed picture [Schna-el, private communication], however, additional investigation iseeded to further clarify the role of defects on electrical propertiesf the B doped Si-NC:SiC material.

. Summary and conclusions

We reported the electrical properties of silicon rich carbide-ased, boron doped multilayers deposited by PECVD, annealed at100 ◦C, fabricated with the aim of obtaining silicon nanocrys-als for photovoltaic applications. The crystallization was checkedy XRD. The technique indicates about 6 nm silicon nanodotiameter, the actual average size being possibly smaller. Optical&T measurements show that the silicon crystallized fraction isubstrate-dependent, and higher values are observed on c-Si ratherhan on quartz substrate. FTIR transmittance shows that remark-ble, although not complete, boron interdiffusion occurs betweendjacent layers, however, the material is shown to keep mem-ry on the initial distribution of boron. Electrical measurementsive indication of the presence of an unwanted doping level inhe crystallized SiC, possibly due to oxygen or nitrogen, which isn analogy with what reported in the past for PECVD deposited

icrocrystalline silicon. Boron doping acts as compensator and-type dopant. When introduced in the multilayer, there is evi-ence of preferential location of B at the interface of Si relatedefects or grain boundaries. Boron doping in the average produces

Please cite this article in press as: C. Summonte, et al., Mater. Sci. Eng. B (2

n improvement of the electronic properties of the SiC material. Theigh room temperature conductivity obtained reflects the limitedand gap of SiC as dielectric matrix, thus making it suitable forpplication in photovoltaic devices.

[

[

-Si, (c-SiC, aSi, aSiC, SCR, and intrinsic multilayer. (b) Boron doped single layer SiC

Acknowledgements

The research leading to these results has received fundingfrom the European Community’s FP7 (FP7/2007-2013) under grantagreement n◦. 245977 under the project title NASCEnT. One of theauthors (Rimpy Shukla) undertook this work with the support ofthe ICTP (International Center for Theoretical Physics) Programmefor Training and Research in Italian Laboratories, Trieste, Italy.

References

[1] G. Conibeer, M. Green, E.C. Cho, D. König, Y.H. Cho, T. Fangsuwannarak, G.Scardera, E. Pink, Y. Huang, T. Puzzer, S. Huang, D. Song, C. Flynn, S. Park, X.Hao, D. Mansfield, Thin Solid Films 516 (2008) 6748–6756.

[2] C.W. Jiang, M.A. Green, Journal of Applied Physics 99 (2006) 114902.[3] S. Godefroo, M. Hayne, M. Jivanescu, A. Stesmans, M. Zacharias, O.I. Lebedev, G.

van Tendeloo, V.V. Moshchalkov, Nature Nanotechnology 3 (2008) 174–178.[4] A.M. Hartel, S. Gutsch, D. Hiller, M. Zacharias, Physical Review B 85 (2012)

165306.[5] S. Ossicini, M. Amato, R. Guerra, M. Palummo, O. Pulci, Nanoscale Research

Letters 5 (2010) 1637–1649.[6] M. Zacharias, J. Heitmann, R. Scholz, U. Kahler, M. Schmidt, J. Bläsing, Applied

Physics Letters 80 (2002) 661–663.[7] I. Perez-Wurfl, X. Hao, A. Gentle, D.-H. Kim, G. Conibeer, M.A. Green, Applied

Physics Letters 95 (2009) 153506.[8] I. Perez-Wurfl, L. Ma, D. Lin, X. Hao, M.A. Green, G. Conibeer, Solar Energy

Materials and Solar Cells 100 (2012) 65–68.[9] D. Song, E.C. Cho, G. Conibeer, C. Flynn, Y. Huang, M.A. Green, Solar Energy

Materials and Solar Cells 92 (2008) 474–481.10] X.J. Hao, E.C. Cho, C. Flynn, Y.S. Shen, S.C. Park, G. Conibeer, M.A. Green, Solar

Energy Materials and Solar Cells 93 (2009) 273–279.11] R. Lechner, A.R. Stegner, R.N. Pereira, R. Dietmueller, M.S. Brandt, A. Ebbers,

M. Trocha, H. Wiggers, M. Stutzmann, Journal of Applied Physics 104 (2008)053701.

12] S.H. Hong, Y.S. Kim, W. Lee, Y.H. Kim, J.Y. Song, J.S. Jang, J.H. Park, S.H. Choi, K.J.Kim, Nanotechnology 22 (2011) 425203.

13] D. Cavalcoli, F. Detto, M. Rossi, A. Tomasi, A. Cavallini, Nanotechnology 21 (2010)045702.

14] I. Balberg, Journal of Applied Physics 110 (2011) 061301.15] G.R. Chang, F. Ma, D.Y. Ma, K.W. Xu, Nanotechnology 21 (2010) 465605.16] R. Gradmann, P. Loeper, M. Künle, M. Rothfelder, S. Janz, M. Hermle, S. Glunz,

Physica Status Solidi C 8 (2011) 831–834.17] Y. Kurokawa, S. Tomita, S. Miyajima, A. Yamada, M. Konagai, Proc. 24th

European Photovoltaic Solar Energy Conference, Hamburg, Germany, 21–25September, 2009, pp. 398–403.

18] P. Löper, D. Stüwe, M. Künle, M. Bivour, C. Reichel, R. Neubauer, M. Schnabel, M.Hermle, O. Eibl, S. Janz, M. Zacharias, S.W. Glunz, Advanced Materials 24 (2012)3124–3129.

19] R. Shukla, C. Summonte, M. Canino, M. Allegrezza, M. Bellettato, A. Desalvo, D.Nobili, S. Mirabella, N. Sharma, M. Jangir, I.P. Jain, Advanced Materials Letters

012), http://dx.doi.org/10.1016/j.mseb.2012.10.030

3 (2012) 297–304.20] F. Gourbilleau, C. Ternon, D. Maestre, O. Palais, C. Dufour, Journal of Applied

Physics 106 (2009) 013501.21] H. Rinnert, M. Vergnat, A. Burneau, Journal of Applied Physics 89 (2001)

237–243.

ING Model

M

8 ce and

[

[

[[

[[

[

[

[

[

[

[

[

[

[

ARTICLESB-13241; No. of Pages 8

C. Summonte et al. / Materials Scien

22] C. Summonte, M. Canino, M. Allegrezza, M. Bellettato, A. Desalvo, S. Mirabella, A.Terrasi, Proc. 26th European Photovoltaic Solar Energy Conference, Hamburg,Germany, 5–8 September, 2011, pp. 361–366.

23] M. Canino, C. Summonte, M. Allegrezza, R. Shukla, M. Bellettato, A. Desalvo, F.Mancarella, M. Sanmartin, A. Terrasi, P. Löper, M. Schnabel, S. Janz, Presentedat the E-MRS 2012 Spring Meeting, Symp. Y, Strasbourg, France, May 13–18,2012, Materials Science and Engineering B, submitted for publication.

24] D.E. Aspnes, A.A. Studna, E. Kinsbron, Physical Review B 29 (1984) 768–779.25] G.E. Jellison, M.F. Chisholm, S.M. Gorbatkin, Applied Physics Letters 62 (1993)

3348–3350.26] G.E. Jellison, F.A. Modine, Applied Physics Letters 69 (1996) 371–373.27] T. Syau, B.J. Baliga, R.W. Hamaker, Journal of the Electrochemical Society 138

(1991) 3076–3080.28] E. Agocs, P. Petrik, S. Milita, L. Vanzetti, S. Gardelis, A.G. Nassiopoulou, G. Pucker,

Please cite this article in press as: C. Summonte, et al., Mater. Sci. Eng. B (2

R. Balboni, M. Fried, Thin Solid Films 519 (2011) 3002–3005.29] S. Gardelis, A.G. Nassiopoulou, P. Manousiadis, S. Milita, A. Gkanatsiou, N.

Frangis, C.B. Lioutas, Journal of Applied Physics 111 (2012) 083536.30] M. Canino, M. Allegrezza, R. Shukla, M. Bellettato, A. Desalvo, M. Perani, D.

Cavalcoli, I.P. Jain, I. Crupi, A. Terrasi, S. Mirabella, C. Summonte, Presented at

[

[

PRESS Engineering B xxx (2012) xxx– xxx

the 27th European Photovoltaic Solar Energy Conference, Frankfurt, Germany,24–28 September, 2012.

31] P. Scherrer, Nachrichten von der Gesellschaft der Wissenschaften zu Göttingen,Mathematisch-Physikalische Klasse, Band (1918) 98–100.

32] E. Centurioni, Applied Optics 44 (2005) 7532–7539, see also:http://www.raysolar.com/optical/optical.html

33] J.L. Prince, F.N. Schwettman, Journal of the Electrochemical Society 121 (1974)705–710.

34] P. Torres, J. Meier, R. Flückiger, U. Kroll, J.A. Anna Selvan, H. Keppner, A. Shah,S.D. Littelwood, I.E. Kelly, P. Giannoulès, Applied Physics Letters 69 (1996)1373–1375.

35] J. Meier, R. Flückiger, H. Keppner, A. Shah, Applied Physics Letters 65 (1994)860–862.

36] R. Guerra, I. Marri, R. Magri, L. Martin-Samos, O. Pulci, E. Degoli, S. Ossicini,

012), http://dx.doi.org/10.1016/j.mseb.2012.10.030

Physical Review B 79 (2009) 155320.37] Y. Kurokawa, S. Yamada, S. Miyajima, A. Yamada, M. Konagai, Current Applied

Physics 10 (2010) S435–S438.38] D. Azulay, O. Millo, E. Savir, J.P. Conde, I. Balberg, Physical Review B 80 (2009)

245312.