Marys Medicine

Depth dependent properties of ito thin films grown by pulsed dc sputtering

ARTICLE IN PRESS
Contents lists available at Materials Science and Engineering B Depth dependent properties of ITO thin films grown by pulsed DC sputtering A. Sytchkova , D. Zola , L.R. Bailey , B. Mackenzie , G. Proudfoot , M. Tian , A. Ulyashin a ENEA Optical Coatings Laboratory, via Anguillarese 301, 00123 Rome, Italy b Oxford Instruments Plasma Technology, Yatton, Bristol, BS49 4AP, UK c NT-MDT Europe BV, High Tech Campus 83, 5656 AG Eindhoven, The Netherlands d SINTEF Materials and Chemistry, Forskningsveien 1, P.O. 124 Blindern, NO-0314 Oslo, Norway A systematically prepared set of ITO layers for solar cell applications has been analyzed by spectroscopic Received 7 July 2012 variable angle ellipsometry in order to trace the dependence of free carriers' distribution along the film Received in revised form 13 October 2012 depth as a function of film thickness as well as its change upon annealing. Samples were deposited on Accepted 25 November 2012 silicon substrates with various thicknesses in steps of approximately 10–20 nm. This set was duplicated Available online xxx and these samples were annealed, so that for each thickness an as-deposited and an annealed sample is available. Conventionally measured electrical conductivity and morphological properties (AFM mea- surements) of the films have been compared with the optical constants' inhomogeneity, i.e. material properties along the film thickness modelled by variable-angle spectroscopic ellipsometry. The obtained results show that the optical as well as electrical properties of thin ITO films prepared by pulsed DC sputtering are depth dependent. For the deposition conditions used a well-determined reproducible non-uniform distribution of free carriers within the film thickness was determined. In particular it has been found that the majority of free carriers in as-deposited ultra-thin ITO films is concentrated at sample half-depth, while their distribution becomes asymmetric for the thicker films, with a maximum located at approximately 40 nm depth. The distribution of free carriers in annealed samples is qualitatively different from that of as-deposited layers.
2012 Elsevier B.V. All rights reserved.
thermoionic emission of carriers across the grain boundaries While scattering from grain boundaries and ionized impurities is Indium tin oxide (ITO) continues to be the material of choice the dominant free carriers' mobility limitation in materials with for many applications where highly conductive thin films of high relatively low lattice mobility, for ITO these two scattering pro- transparency are required: flat panel displays, solar cells, defrost- cesses should be accomplished by other scattering mechanisms, ing windows, electromagnetic shielding etc. is one of the e.g. for heavily doped semiconductors like 10%- most studied transparent conducting oxides (TCOs) in terms of one deals with a mixture of phases with different dependencies of its optical, electrical, morphological and structural conductivities, whereas in the case of a thin film a contribution from properties on deposition conditions by various methods (see for different structural material organizations, amorphous and crys- example the history of processes for making TCOs) as well talline, should be also considered these factors determine as on post-deposition treatment like annealing, e.g. plasma the overall trap density that is the measure of TCO electrical per- hydrogenation combination of good optical and formance. Recently, for polycrystalline magnetron sputtered ITO, electrical performance of this material may be properly tuned, a significant variation of the mobilities in the carrier concentra- according to an application demands, by variation of deposition tion range N > 3×1020 cm−3 was found and a phenomenological conditions and by the subsequent treatment.
model was created for temperature-dependent mobility of carriers Carrier transport in TCOs have been investigated in a series of that combines a temperature-independent term with a metal-like works modelling the conductivity mainly based on the Hall and contribution, i.e. a thermally activated part due to grain boundary- Seebeck measurements, e.g. provide the values for carrier limited transport In the cited work an attempt was done concentration and mobility in the film. Such modelling combines moreover to find a correlation between the discharge voltages compensated electron conduction within the film crystallites and and the trap densities for different deposition configurations of sputtering, including DC or RF excitations and different RF frequen- cies. While such a correlation was established for the other of two studied materials, no correlation was found for ITO, despite the ∗ Corresponding author.
well-known significant influence of reactive oxygen on the TCO E-mail address: (A. Sytchkova).
growth and properties, e.g. 0921-5107/$ – see front matter 2012 Elsevier B.V. All rights reserved.
Please cite this article in press as: A. Sytchkova, et al., Mater. Sci. Eng. B (2012), ARTICLE IN PRESS
A. Sytchkova et al. / Materials Science and Engineering B xxx (2012) xxx–xxx Deposition time for the pulsed DC sputtered ITO films and their electrical properties and ellipsometrically simulated thickness (surface roughness overlayer is not included).
Resistivity (␮ cm) Resistivity after anneal Film thickness, ellips The values of carrier concentration and mobility are character- ellipsometric data is shown to be a reliable indicator of the film istics of the film as a whole, i.e. in the mentioned models the film roughness obtained by the AFM technique.
material is considered homogenous. Little is known about the inter- nal organization of free carriers within the film thickness and in 2. Experimental details
general about thickness dependence of ITO optical and electrical properties. Information on the carrier distribution within the film The samples were deposited on one-side polished crystalline depth is useful for several reasons like, for example, adjustment 100 p-doped silicon substrates using a demonstration pulsed DC of the deposition process parameters or those of post-deposition sputtering tool PlasmaPro System 400 (Oxford Instruments Co.).
treatment, as well as minimization of the material use once an opti- Depositions were carried out in a reactive gas chemistry and pre- mal film thickness with a sufficient number of charge carriers is cision controlled O from a 200 mm diameter target of 99.99%-pure ITO (SnO2:In2O3 10:90% wt) equipped with a mag- A study on the microstructure and electrical properties thick- netron, 290 W DC power with a pulse frequency of 100 kHz. The ness dependence for ITO film deposited by RF magnetron sputtering processing pressure at the start and during deposition was 2 m Torr.
on plastic substrate is reported in Ref. The study was per- The percentage of oxygen to argon was 2%. From one deposition run formed on a set of samples with thicknesses varying in the to another only the film thickness, that is the run time, was var- range 40–280 nm (measured by a profilometer), and changes in ied with a step of 12 s which corresponds to about 10 nm. Half of microstructural and morphologic properties were observed using the samples were then annealed at 160◦ for 2 h in air. Our research a scanning electron microscope. No optical characterization of the found that annealing ITO at 160 ◦C for 2 h was equivalent to anneal- film properties was performed, or any study of carrier concentra- ing at 300 ◦C for 1 h. The lower temperature anneal was utilized tion versus film thickness variation.
since there are applications where device damage could result Recently, an ellipsometric study of ITO inhomogeneity was in using higher temperatures. the deposition performed for the RF sputtered films of approximately 1 ␮m thick- times and reports the film resistance values obtained by van der ness geometrical variation of the refractive index of the Pauw four-point technique on twin samples deposited during the material, and in particular its imaginary part, i.e. the extinction same runs on glass.
coefficient, was noticed as an indicator of free carrier distribu- The samples were measured with a variable angle spectroscopic tion along the film depth. The effects of deposition temperature ellipsometer (WVASE, J.A. Woollam Co.) in the spectral range of and post-deposition annealing on re-distribution of the carriers 300–2500 nm. At least three incidence angles were used for this were traced as well. Moreover, the cited study has underlined analysis, 55◦, 65◦ and 75◦, while the thinnest layers were addition- the importance of the film inhomogeneity modelling for a correct ally checked at a grazing angle of 85◦.
determination of the film thickness. In fact, inadequate character- All the samples were investigated by atomic force microscopy ization yielding imprecise film thickness values preclude correct (AFM) on an Ntegra integrated system (NT-MTD Co.). AFM mea- determination of charge carrier density, mobility, the effective elec- surements were carried out in semi-contact mode using a standard tron mass, the scattering mechanisms and relaxation time silicon probe. The probes were used with a spring constant of imprecision in film thickness value determination may be either ∼2.5 N/m and a free resonance of ∼150 kHz. The silicon tip at due to a poor accuracy of the measurement technique, like that of the end of these AFM probes had a curvature of ∼10 nm. Both mechanical profilometry, or due to deficient modelling of the indi- topography and phase images were recorded simultaneously in the rect measurements like modelling of optical spectra without film scanning areas of 5 ␮m × 5 ␮m and 1 ␮m × 1 ␮m. All the images were collected in air at scan rate of 1.0 Hz with 600 × 600 pixels In this work we present a methodical study of optical, electrical resolution. AFM images were analyzed by using Nova Px software and morphological properties of ITO performed on a set of layers (NT-MDT Co.) yielding surface roughness parameters of ‘Root Mean with thickness varying from 10 to 80 nm, the thickness mostly use- Square (RMS) roughness' (Rq) and average roughness (Ra).
ful for solar cell applications. The insight into film properties change with its growth is achieved thanks to observation of a systematic 3. Ellipsometric analysis
law of the film refractive index inhomogeneity studied together with the thickness dependence of the layer morphology and resis- The ITO material optical dispersion was described by a gen- tivity. Moreover, film surface modelling by a porous overlayer for eral oscillator model containing three different oscillators: one Please cite this article in press as: A. Sytchkova, et al., Mater. Sci. Eng. B (2012), ARTICLE IN PRESS
A. Sytchkova et al. / Materials Science and Engineering B xxx (2012) xxx–xxx Tauc-Cauchy that is representative for the UV–vis part of the Overlayer with 50% voids acquired spectrum and two Lorentz ones which express the NIR optical behaviour of the material. The two Lorentz oscillators were necessary to accurately reproduce the shape of the curves of the SiO native oxide 1.63 nm ellipsometric angles, most probably due to a slight asymmetry Si substrate 0.5 mm of the absorption band in the NIR. In fact one of them is much less intensive and broader than the other. Notice that use of an Fig. 1. Schematic presentation of the layered model for ellipsometric characteriza-
extended Drude model (EDM) is another valid way for describing tion of the ITO/Si samples.
the ITO film transport processes particular because it accounts for ionized impurity scattering. Typically, EDM is used correspond to any physical process of film growth, while from a to correct Drude approximation for interband transitions, which mathematical point of view this would add one more fit parame- in practice means additional Lorentz-type oscillator(s) into the ter and hence, increase instability of the inverse problem of film original Drude model. The choice of the dispersion model should characterization. Both the refractive index n and the extinction not however influence the conclusions on film inhomogeneity as coefficient k of the film are functions that may be considered a each of the models leads to a reasonable fit of experimental data product of their dispersion and geometrical dependencies: within reliable values of the optical constants.
n(, z) = n()q(z); k(, z) = k()q(z) The overall dispersion of the material was then considered smoothly variable along the layer depth using a three-pole grad- where  is the test wavelength and 0 < q(z) ≤ 1 is the film profile ing. A higher polynomial order was inefficient as it would not function describing the optical functions' variation within the film Fig. 2. Dispersion of the optical constants for as-deposited (a) and annealed (b) ITO samples.
Distance from Substrate in nm Distance from Substrate in nm Distance from Substrate in nm Distance from Substrate in nm Fig. 3. Optical constants profiles for the thickest among as-deposited (256, left) and annealed samples (241, right): at 500 nm (a and c) and at 1500 nm (b and d).
Please cite this article in press as: A. Sytchkova, et al., Mater. Sci. Eng. B (2012), ARTICLE IN PRESS
A. Sytchkova et al. / Materials Science and Engineering B xxx (2012) xxx–xxx clearly seen that the annealing treatment has provoked a blue shift of both the energy gap and the plasma frequency of the material. For further analysis of the sample set it is worth noting right now that there is no physical reason to expect optical dis- persion laws different from these two for other samples of each category. For this reason these laws were kept invariable so that only the film and its overlayer thicknesses were fitting parameters together with the geometrical profiles of the optical constants. The geometrical profile of the optical constants of the ∼80 nm thick samples are shown in two wavelengths representative for the visible and NIR part of the spectrum. It is rather compre- hensible that at 500 nm ITO behaves like a dielectric so no traces of free carrier distribution within the thickness are expectable.
In the vicinity of plasma frequency the situation changes; that is why all the significant for conductivity profiles of k are taken at An accurate profiling of the optical constants not only allows the insight of optical and electrical properties of the material under study, but also ensures the correct film thickness determination.
It is especially true in the case of ultrathin layers. For the mate- rials with a complex dispersion like ITO has, such a profiling is possible only when the measurement range is sufficiently wide to include the region around the material plasma frequency. This may be illustrated using the example of the sample 250 (about 20 nm thick as estimated from the deposition rate and run time). Equally good fitting may result for a restricted range 300–800 nm and for the whole range up to 2500 nm, despite the fact that the first was obtained using invariable-in-the-depth optical constants, while the second was possible only when applying a parabolic depth profiling, The correctness of the latter is proved by the values of the film and overlayer thicknesses. In fact, in the first case the resulting film thickness is about 17.3 nm with Model Fit Exp 55° an overlayer of 3.6 nm, while in the second case: the film thick- ness is 18.2 nm, while the overlayer is just 2.6 nm thick. These last values are better confirmed by the AFM imaging of the sam- ple, see Section as far as the film is a continuous amorphous layer with a low RMS roughness. In addition, another confirma- tion of the second model's accuracy is the fact that a full-range fit without refractive index variation in depth included in the model results in a high discrepancy both in the UV and especially in the NIR range, Corresponding to this last model val- ues of the film and overlayer thicknesses are 15.6 nm and 6.3 nm, It should be noted that in any case the initial approximation for the fit procedure for the ultrathin layers (d < 30 nm) is cru- Fig. 4. Fit of the ellipsometric angle  : (a) short range 300–800 nm within a model
cial. It is important to fix the material dispersion type within the of a constantover-thickness optical constants; (b) full range 300–2500 nm within a model of inhomogeneous optical constants and (c) full range 300–2500 nm within model otherwise meaningless values of the layer and overlayer a model of a constantover-thickness optical constants.
thicknesses will most probably be obtained in a free parameter depth z, when z varies from 0 (substrate) to d (outer surface of the 4. Results and discussions
film). Notice that a linear profile of the optical constants is a possible particular case of our generally asymmetric parabolic description Correct determination of the thickness of the layers thanks to the elaborated approach for ellipsometric characterization of the Before characterization of the ITO layers, the silicon substrate samples has allowed precise calculation of their resistivity, see last with its native oxide was measured and the oxide thickness was column of film electrical and morphological properties estimated together with dielectric functions of the silicon. The ITO were compared with the optical constants inhomogeneity within layer roughness was modelled using a surface overlayer with 50% the film thickness.
void inclusions sketch of the layered model is shown in the extinction coefficient profiles for the samples The two thickest of the layers allow the most reliable in-depth grown on Si substrates, taken at the wavelength of 1500 nm, which modelling thanks to higher ratio d/, where d is the film thick- is representative for the conducting properties of the ITO material, ness. Thus, we used them as a starting point for characterization i.e. for mobile carriers distribution within the film. It is possible to of the whole sample set and let us consider first these samples deduce that the ITO film formation on Si ensures a certain non- as examples of ellipsometric data fit procedure. dis- uniform distribution of concentration of mobile electrons within persion of the optical constants of the ∼80 nm thick films. It is the film thickness: for the deposition process under study, the Please cite this article in press as: A. Sytchkova, et al., Mater. Sci. Eng. B (2012), ARTICLE IN PRESS
A. Sytchkova et al. / Materials Science and Engineering B xxx (2012) xxx–xxx series of ITO thin films. Obviously, more microstructural features with larger grains formed on the ITO thin films with longer depo- sition time and thermal annealing. The grains are developed when the thickness of ITO thin film is above 50 nm (samples 244–241 and 253–256). No observable grains even at largest scale scans should mean that we may consider such films amorphous or at least hav- ing prevailing amorphous component (absence of tiny crystals of a few nm-size cannot be proven without structural characteriza- tion, of course). This conclusion is not surprising if consider that the samples were deposited without substrate heating. When com- paring the morphology images for ITO thin films of samples 249, 254, and 256, larger ITO grains are observed when ITO thickness increases; actually, the ITO lamellae crystals are investigated on the AFM images of the thickest films (samples 241 and 256). This indicates that the larger and well-organized crystal forms when increasing ITO thickness, and crystalline degree increases as well Fig. 5. Thicknesses and extinction coefficient profiles of as-deposited ITO samples.
upon film annealing. Another indirect proof of the film structure type is the maintenance of the shape of the extinction coefficient profiles upon annealing for the films thinner than 50 nm: sam- ples 248–246 have same symmetric k-profile like as-deposited majority of free carriers and/or those with high mobility in as- samples, and the profiles start loose the symmetry starting from deposited ultra-thin ITO films is concentrated at sample half-depth, sample 245. This must mean that those films were amorphous while their distribution becomes asymmetric for the thicker films, when deposited and did not undergo crystallization upon anneal- with a maximum located at approximately 40 nm depth. Moreover, ing, while sample 245 is presumed to possess a mixture of phases.
this behaviour is reproducible from sample to sample. Notice that On the other hand, the films having grains after deposition evolved the samples were deposited at low powers for a maximum of 1 min into more crystalline form confirmed by AFM and by k-profile and 36 s so therefore the substrate heating during deposition was Based on these topography images, surface roughness analy- For the annealed samples, the material distribution within the sis was done with yielding root mean square (RMS) and average film depth changes and becomes far more asymmetric, with extinc- roughness values. the main roughness parameters tion coefficient well-pronounced maximum located in the vicinity of both as-deposited and annealed ITO films on silicon in com- of the substrate: the main concentration and mobility of the free parison with ellipsometrically modelled overlayer thickness. It is carriers in ITO film is within the first 10 nm of the film, see possible to observe a correlation in these parameters, but the abso- Annealing is known to improve ITO performance thanks to dimin- lute values for surface roughness characterization obtained by AFM ished disorders and enhanced crystallinity of the material leading and by ellipsometry differ significantly. The main reason for this is to improved electron mobility. A hypothesis on a non-uniform crys- that optical probing of the surface in the UV–NIR range reveals the tal growth and lattice adjustment during annealing may be done.
roughness originated by surface features with spatial frequencies Such an effect occurs for the layers with thicknesses over 40 nm which are very different from those AFM refers to. Moreover, mod- (with sample 245 showing a "threshold" behaviour), while for thin- elling of the overlayer as a 50% Bruggeman-type mixture of air voids ner films amorphous structure continues to prevail, as confirmed and the underlying ITO material is a large-scale approximation over also by AFM measurements. This non-uniformity should follow a much bigger surface integration area. Nevertheless, the trend of the temperature distribution inside the film when the substrate the overlayer thickness respects the RMS and average roughness is heated, while the outer part of the film results more oxidized values detected by AFM. It is thus concluded that the film rough- compared to its internal fraction.
ness can be correctly monitored by ellipsometry, even for ultrathin shows AFM topography images for the thickest, the thinnest and an intermediate of both as-deposited and annealed Electrical resistivity ott at optical frequencies may be retrieved from the optical parameters of the film considering that for the optical range when n ≈ k, at each position z along the film thickness, the conductivity (z) of the considered sub-layer is related to its extinction coefficient as following where c is the light velocity in vacuum and ε0 is the electric permit- tivity of vacuum. Therefore, the total film conductivity  calculated if consider the parallel connection of all the sub-layers along the film depth: Fig. 6. Thicknesses and extinction coefficient profiles of annealed ITO samples.
Please cite this article in press as: A. Sytchkova, et al., Mater. Sci. Eng. B (2012),

ARTICLE IN PRESS
A. Sytchkova et al. / Materials Science and Engineering B xxx (2012) xxx–xxx Fig. 7. Topography 1 ␮m × 1 ␮m of selected ITO layers on Silicon: as deposited (left) and annealed (right).
The calculated "optical resistivity" for both as-deposited and difference in thickness will result in a disproportionate change in annealed films is shown in comparison with their DC resis- the measured resistivity. Anyway, sample 248 is estimated about 15% thicker than sample 249 what confirms the tendency in their 0 measured by the four-probe method. The two types of resistivity have similar behaviour for both sample sets, and in addi- resistivity values.
tion the minimum value of resistivity is obtained for the annealed Notice that the values of 0 and ott must be differ- 40 nm thick sample (sample 245) using these two independent ent because different frequencies are considered. However, if measurement techniques. The minimum at 40–50 nm is proba- Drude-like conductivity acts in our samples, the trend of the bly connected with the phase mixture nature of these samples two resistivities should be very similar otherwise it would (crystalline and amorphous). Sample 247 exhibits no change in the mean that non-optimal doping depresses the Drude-like con- refractive index with annealing and since this implies there is no ductivity favouring other transport mechanisms caused, for change in the crystal structures it is not unreasonable for there to example, by the presence of amorphous regions, localized states be no change in the resistivity, see The samples 248 and etc. These observations corroborate that the optical extinc- 249 evidently have significantly different values of resistivity. How- tion function is a measure of electrical properties of the ITO ever, they are both thin ∼10 nm and as is well known, any slight Please cite this article in press as: A. Sytchkova, et al., Mater. Sci. Eng. B (2012),

ARTICLE IN PRESS
A. Sytchkova et al. / Materials Science and Engineering B xxx (2012) xxx–xxx RMS and average roughness values from AFM measurements on ITO film versus ellipsometrically determined surface overlayer.
Roughness average Roughness average thickness. Detection of re-distribution of the carriers upon anneal- ing was another result of the VASE data modelling.
Finally, this type of analysis has allowed, for the first time to the best of our best knowledge, a systematic study of thickness dependent properties of DC-pulsed sputtered ITO films by a non- destructive optical technique.
This work was performed within the frame of EC project ThinSi, grant agreement no. 241281.
neda, Materials Sciences and Applications 2 (2011) 1233–1242.
[2] R.G. Gordon, MRS Bulletin 25 (2000) 52–57.
[3] S. Calan, A.N. Tiwari, Thin Solid Films 518 (2010) 1839–1841.
[4] H.L. Hartnagel, A.L. Das, A.K. Jain, C. Jagadish, Semiconducting Transparent Thin Films, Institute of Physics Publishing, Bristol, 1995.
Fig. 8. Thickness dependence of the electric resistivity 
0 and "optical resistivity" [5] T. Koida, H. Fujiwara, M. Kondo, Journal of Non-Crystalline Solids 354 (2008) ott at 2000 nm of the as-deposited and annealed films.
[6] K. Ellmer, R. Mientus, Thin Solid Films 516 (2008) 4620–4627.
[7] J. Tate, M.K. Jayaraj, A.D. Dareseke, T. Ulbrich, A.W. Sleight, K.A. Vanaja, R.
Nagarajan, J.F. Wager, Thin Solid Films 411 (2002) 119.
[8] K.-S. Weissenrieder, J. Mueller, Thin Solid Films 300 (1997) 30–41.
[9] K. Ellmer, R. Mientus, Thin Solid Films 516 (2008) 5829–5835.
Film inhomogeneity plays an essential role in ITO characteri- [10] D.C. Paine, T. Whitson, D. Janiac, R. Beresford, C.O. Yang, B. Lewis, Journal of zation allowing correct determination of the film thickness and Applied Physics 85 (1999) 8445.
[11] A. Krasilnikova Sytchkova, M.L. Grilli, S. Boycheva, A. Piegari, Applied Physics its optical constants. This is especially true at nanoscale, i.e. for A: Materials Science and Processing 89 (2007) 63–72.
very thin layers (thinner than ∼20 nm). Moreover, free carriers [12] D.-H. Kim, M.-R. Park, H.-J. Lee, G.H. Lee, Applied Surface Science 253 (2006) distribution and mobility along the film depth comprehensible [13] A. Krasilnikova Sytchkova, S. Schutzmann, Bulk and Surface Inhomogeneity of from the shape of the extinction coefficient curve is important for RF Sputtered ITO Films: an Ellipsometric Study in DGaO-PROCEEDINGS, 2009 adjustment of the deposition process parameters or those of post- [ISSN: 1614-8436] deposition treatment, as well as minimization of the material use [14] R.J. Mendelsberg, G. Garcia, J.D. Milliron, Journal of Applied Physics 111 (2012), once an optimal film thickness with a sufficient number of charge [15] M. Losurdo, M. Giangregorio, P. Capezzuto, G. Bruno, R. De Rosa, F. Roca, C.
carriers is found.
Summonte, J. Plá, R. Rizzoli, Journal of Vacuum Science and Technology A 20 Film roughness can be correctly monitored by ellipsometry, (2002) 37–42.
even for ultrathin layers. Nanoscale effects may be successfully [16] K. Füchsel, U. Schulz, N. Kaiser, A. Tünnermann, Applied Optics 47 (2008) studied by VASE, when the measurement covers the range of essen- [17] A. Tikhonravov, M. Trubetskov, A. Krasilnikova, E. Masetti, A. Duparré, E. Ques- tial features of the material and a proper modelling methodology nel, D. Ristau, Thin Solid Films 397 (2001) 229–237.
is elaborated.
[18] (a) F. Bassani, G. Pastore Parravicini, Electronic State and Optical Transition, Pergamon Press, Oxford, 1975; The ITO film formation on Si ensures a sample to sample (b) G. Grosso, G. Pastore Parravicini, Solid State Physics, Academic Press Inc., reproducible non-uniform distribution of carriers within the film San Diego, 2000.
Please cite this article in press as: A. Sytchkova, et al., Mater. Sci. Eng. B (2012),

Source: http://sintef.se/project/ThinSi/pdf/Materials%20Science%20and%20Engineering%20B_2013.pdf