Structural Origin of Magnetotransport Properties in APCVD Deposited Single and Bi-Layer Tin Oxide Thin Films

Tin oxide (SnO2) thin films, undoped single-layer and fluorine doped / undoped bilayer were deposited by Atmospheric Pressure Chemical Vapour Deposition (APCVD) at soda-lime glass substrate at two different temperatures (590°C and 610°C). Transport properties examined by impedance spectroscopy, DC resistivity, Hall effect and magnetoresistance measurements are correlated with structural properties examined by Scanning Electron Microscopy, Grazing Incidence X-ray Diffraction (GIXRD) and Time-of-flight Elastic Recoil Detection Analysis (TOF-ERDA). Results suggest that charge transport in the obtained samples is dominated by scattering at neutral impurities and can be correlated with preferred orientation (texture coefficient) presented in prepared samples.


Introduction
Transparent conductive oxides (TCO) are binary or ternary compounds containing one or two metallic elements. TCO materials are characterized by a very good balance of optical and electrical properties. Widely used TCO are oxides such as ZnO, SnO2, In2O3 doped with metallic elements: Al-doped ZnO (AZO), Sn-doped In2O3 (ITO) and F-doped SnO2 (FTO) [1][2][3].
Among others, tin oxide (SnO2), also known as stannic oxide, is an n-type semiconductor (due to oxygen vacancies) with high optical transparency in visible spectral range (> 85%) and wide energy band gap (3.6 eV). It can be found in nature as a mineral known as cassiterite, and it is the main ore of tin [4]. It has a rutile-like crystal structure. The SnO2 thin films are chemically inert, scratch resistant, and can withstand high temperatures [5].
SnO2 thin films (pure and doped) have various applications in devices used in daily life: solar energy conversion, flat panel displays, electro-chromic devices, invisible security circuits, LEDs, transparent electrical conductors and non-colouring electrodes, in smart windows, for energy and illumination control, in anti-dazzling rear view windows, and non-emissive displays, low-emittance coatings for energy efficient windows, anti-frost coatings on car windows and transparent electrode for solar cells. In these applications the film thickness normally lies in the 100-1000 nm range [5].
For photovoltaic application (as transparent front electrode) it is important that SnO2 is transparent for UV-VIS light and has very good conductivity at the same time. Electrical conductivity can be improved by increasing charge carrier density (doping with foreign atoms) or charge carrier mobility. The most favourable dopants are antimony (SnO2:Sb) and fluorine (SnO2:F). Fluorine doped tin oxide (SnO2:F, FTO) exhibits good visible transparency owing to its wide band gap, while retaining a low electrical resistivity due to the high carrier concentration caused by the oxygen vacancies and the fluorine dopant. Higher number of charge carriers causes lower film transparency. Therefore, better conductivity should be achieved by optimizing carrier mobility. Several different approaches are reported how to improve mobility of SnO2 such as post-deposition heat treatment, deposition technique, substrate, doping control [7]. For example, mobility can be improved by use of highly oriented substrate, use of tin tetrachloride as Sn precursor, higher methanol content.
There are various chemical and physical methods for preparation of pure and doped SnO2 thin films: chemical vapour deposition, sol gel, spray pyrolysis, electron beam evaporation, vapour deposition, pulsed laser deposition, molecular beam epitaxy, thermal evaporation, reactive evaporation and magnetron sputtering, reactive magnetron sputtering, ion beam deposition [5].
In this work Atmospheric Pressure Chemical Vapour Deposition (APCVD) was used for SnO2 thin films deposition. APCVD is a process which enables deposition of vapour species into thin solid films via suitable chemical reactions at atmospheric pressure. By careful choice of the deposition parameters the film properties can be systematically targeted. APCVD is often used in industry for thin film coatings deposition because of reduced costs due to low material consumption, high deposition rates and running costs (compared to low pressure systems) [6].
In our previous publication [8], we reported in detail about structural properties examined by XRD of SnO2 thin film samples deposited by APCVD with a short discussion about the influence on optical (average transmittance in VIS part of the spectrum) and transport properties (specific surface resistivity). In this work, we have investigated in more detail the transport properties of APCVD deposited SnO2 thin films and corelated with structural properties. We present how a very small variation in APCVD deposition temperature results in significantly different transport properties (charge carrier mobility). We explain these differences by variation in structural properties (mainly preferred nanocrystal orientation). For this purpose, we prepared undoped SnO2 thin film single-layer samples and undoped / fluorine-doped SnO2 (SnO2:F) bilayer thin films at two different temperatures (590°C and 610°C). Transport properties were examined in a wide temperature range by impedance spectroscopy, DC resistivity, Hall effect and magnetoresistance. Structure and composition of SnO2 thin films were examined by Scanning Electron Microscopy (SEM), Grazing Incidence X-ray Diffraction (GIXRD) and Time-of-flight Elastic Recoil Detection Analysis (TOF-ERDA).

Thin film deposition by APCVD
SnO2 thin film samples were prepared by the APCVD on the soda-lime glass substrate, in an industrial moving belt reactor [9] with constant-temperature zones. The reactor oven is interrupted with 2 slots perpendicular to the belt moving directions supplied with setup with nozzle that enable the vapour or reactants flow onto the heated glass surface. The reacting gas mixture was SnCl4, H2O, methanol and oxygen for undoped films while for doped films methanol was omitted and the freon gas was added into the vapour mixture. The vapour of reactants was produced in a "bubbler" by passing a carrier gas through the precursors at room temperature (methanol) or moderately heated to 50°C (SnCl4) or 40°C (H2O). The carrier gas for SnCl4, ethanol and H2O was nitrogen. The temperature of glass substrate before and after deposition was 590°C for one series of samples and 610°C for other. The glass substrates were loaded by a belt into the furnace for SnO2 layer deposition. Single-layer deposition duration of 1.5 minutes and post-deposition thermal treatment of some 30 minutes were adjusted by the belt speed. SnO2 film is formed through the very fast reaction of tetrachloride with water vapour and methanol is often added as a moderator. That is why for doped samples where the methanol is omitted, the growth rate is some 30% faster that is similar to the result reported in ref. [10]. As a result, two types of films were prepared. The first type was created in a single-step process and the produced samples were intrinsic. The second type was prepared by depositing the first layer on the glass substrate in the same way as for the first type while the second (top) layer, which was formed from the solution without methanol and with addition of fluorine atoms in the form of freon (Chlorofluorocarbon, CFC), was deposited on the already formed first layer.

Grazing Incidence X-ray Diffraction (GIXRD)
As-deposited films were thoroughly studied by GIXRD, using the synchrotron radiation source. GIXRD was carried out at the MCX beamline [11] (Elettra synchrotron, Trieste, Italy) with the wavelength of the incident beam of 1.5498 Å (8 keV). The angle of incidence was set to 2.0° (value much higher than the critical angle for total external reflection for SnO2 , calculated from SnO2 index of refraction, real part δ [12]). For the critical angle, the beam penetrates 10-20 nm below the surface and gives information about the surface morphology. For the angle of incidence αi = 2.0˚ the beam penetrates much deeper and the GIXRD pattern contains the morphological information for the entire SnO2 layer.

Time-of-flight Elastic Recoil Detection Analysis (TOF-ERDA)
Atomic content and depth profiles of the elements in the samples were determined using TOF-ERDA. TOF-ERDA measurements were done by 20 MeV 127 I 6+ ions with 20° incidence angle toward the sample surface, and TOF-ERDA spectrometer positioned at the angle of 37.5° toward the beam direction. More details about TOF-ERDA setup used in this work can be found in ref. [13].

Transport characterization
2.3.1. Impedance spectroscopy Sheet conductivity of SnO2 thin film samples was measured by impedance spectroscopy (Novocontrol Alpha-N dielectric analyser) in the frequency range from 0,01 Hz to 1 MHz, in three subsequent temperature cycles: i) cooling from 20 C to -100 C, ii) heating from -100 C to 220 C, iii) cooling from 220 C to 20 C. Temperature step for all cycles was 40 C. Sheet conductivity data were normalized to the film thickness (Table 1) to obtain specific conductivity (S/cm).

Magnetotransport
DC resistivity ρ, magnetoresistance and Hall effect measurements were done in the temperature interval from 4 to 300 K and in magnetic fields up to 5 T. Thin film samples were cut for the Hall-bar geometry with typical dimensions (10mm x 2mm x 500nm). Two current contacts and three pairs of Hall contacts were made by applying silver paint directly to the surface of the films. Magnetoresistance and Hall effect were measured simultaneously at fixed temperatures and magnetic field, B, going from -5 to 5 Tesla. Magnetic field was oriented perpendicular to the current through the sample and perpendicular to the surface of the films. Magnetoresistance data was The magnetoresistance was standardly determined as . In Figure 1 SEM images of SnO2 thin film samples are presented. Average grain size was calculated from SEM image line profiles using so called "linear intercept technique" as [14] : (1) where C is a total length of test line used, N is the number of intercepts and M is the SEM image magnification.

Scanning Electron Microscopy
There is a significant difference in surface morphology (grain size and shape) for undoped single-layer samples (Figure 1a and 1b) and bi-layer samples (Figure 1c and 1d). Bi-layer samples have larger grains with sharp edges (pyramidal shape), while single-layer samples have smaller grains with a more rounded shape. The single-layer samples were deposited with addition of methanol just as it was done in experiment reported in refs. [10,15,16] which resulted in smoother films with smaller grains and a higher nucleation density as reported in ref. [15,16]. This effect was addressed to the removal of absorbed HCl which is produced during deposition in reaction of SnCl4 with H2O. Removal of absorbed HCl leads to an increase in number of adsorption sites for SnCl4 and H2O, and consequently an increase in the micro-grain density and simultaneously decrease of grain size.
Higher deposition temperature produces samples with smaller grains (surface roughness) for both single-layer and bi-layer samples. This effect could be a result of competing gas-phase and/or surface reactions, which are a complex function of temperature and composition of reactants [17].   Lattice parameters calculated from peak positions (labelled in Figure 2) are presented. Compared to values from literature (a=0.4737(3) nm and c=0.3186(4) nm) all samples have a larger value of the lattice parameter, a, and equal lattice parameter, c. Variations of lattice parameters are due to the presence of intentional (flour) and non-intentional (chlorine) dopant atoms with different ionic radius which substitute oxygen atoms in the lattice. There is also a significant variation in XRD peak intensity ratio from sample to sample which are also different from values expected from literature. This suggests the presence of preferred orientation (texture) in films. For quantitative analysis the texture coefficient for a selected XRD peak was calculated as intensity ratio of the selected XRD peak and the total sum of all fitted XRD peaks using the formula [19,20]: (2) where I(h,k,l) is the measured relative intensity of the plane (h k l), I0(h k l) is the standard intensity of the plane (h k l) taken from literature (COD database, SnO2 COD ID: 9009082 [18]) and N is the number of reflections included in the calculations. Results for four most intense XRD peaks are presented in Table 2. It is interesting that samples deposited at lower temperature (both single and bilayer) have significantly larger texture coefficient for (110) while bi-layer samples have larger (200) texture coefficient. Nanocrystal size and average maximum microstrain were determined by fitting the data to the standard Williamson-Hall model (Figure 4) [21]:

Grazing Incidence X-ray Diffraction
where βhkl is XRD line width, D is nanocrystal size, K is shape factor (0.94) and λ (=0.154nm) is the wavelength of Cu K-α radiation. Williamson -Hall model assumes that the broadening of XRD pattern is attributed to the nanocrystal size-induced or strain induced broadening. Assuming that the nanocrystal size and strain contributions to the line broadening are independent to each other, the observed line breadth is simply the sum of the two independent contributions given by Eq.   Table 2.
Single-layer samples have smaller nanocrystal size calculated from GIXRD and clearly seen from SEM images. Two possible effects/reasons can be responsible for smaller nanocrystal size in single-layer samples. First is smaller thickness of single-layer samples which assume smaller nanocrystals at the surface according to standard thin film growth model [22]. The second one is variation in the deposition temperature. Samples deposited at the lower temperature have larger average nanocrystal size. Temperature affects a complicated set of chemical reaction during the mixing of reactants in vapour phase and at the substrate surface.

TOF-ERDA
TOF-ERDA was employed for the elemental depth profiling of the S-590 and B-590 samples. Due to the overlapping of Sn and scattered I lines in TOF-E spectra (Figure 4a and 4Figure 4b), only first 1000·10 15 at./cm 2 of sample depth was analysed. Energy spectra belonging to each element were analysed using simulation code Potku [23] (slab analysis) and Monte Carlo (MC) code CORTEO [24]. Calculated depth profiles (Potku analysis) are presented at Figure 5a and 5b for S-590 and B-590 respectively. Since the slab analysis does not take into account detector energy resolution and all other contributions to the total energy spread (energy straggling and multiple scattering of incoming and recoiled ions), derived atomic concentrations were used only as an input data for the MC simulation. Average atomic concentrations, calculated by MC simulations, are listed in the Table  3.  The results in Table 3 confirm that undoped single-layer sample (S-590) does not contain flour, as it is expected.  (1:2), within the measured error, considering that part of O atoms are bonded to Si in substrate or C atoms at the sample surface. For doped bi-layer samples concentration of F atoms (dopant) is almost 1% while for undoped sample the amount of F atoms is below detection limit (<0.1 at. %).
Small amount of Si, K, Mg and Na, visible in TOF-E spectra (Figure 4a and 4b), could originate from the glass substrate since the sample area is not fully covered by SnO2 film. Higher number of holes/cracks is expected for thinner samples (single-layer) which could explain higher contribution of Si, K, Mg and Na in the sample S-590 (Table 3).

Transport properties
3.2.1. Impedance spectroscopy Impedance spectroscopy results ( Figure 6) show that all samples have a very high electrical conductivity, independent of frequency in a wide frequency range indicating fast electronic transport. As expected, bi-layer samples have higher conductivity compared to single-layer deposited at the same temperature because of doping. Only sample S-590 shows a very small dispersion at the highest frequencies. In addition, for all samples the electrode polarisation effects are not observed, indicating the absence of ion transport contribution to the electrical conductivity. Figure 6 Electrical conductivity as a function of frequency at 20ºC for all SnO2 samples.

Magnetotransport probe
DC resistivity (Figure 7) for all samples (single-layer and bi-layer) is practically temperature-independent in a wide temperature range up to the room temperature (4-300K) indicating that the Fermi level is either in the conductive band or in the region where the conductive band is mixed with impurity levels. As well, above room temperature resistivity is independent on temperature and have same trend as below room temperature. Temperature independent resistivity shows that we have a metal type of conductivity, but the dominant scattering is not on phonons but on neutral defects. Namely, only the scattering on neutral impurities is temperature independent, and all other scatterings has the same temperature dependence. Figure 7 clearly shows that undoped samples S-590 and S-610 have greater resistivity than the doped samples B-590 and B-610. This is in line with expectations. There is a small difference in resistivity between the two undoped samples and also between the two bi-layer (doped) samples. This can be explained by slightly different deposition conditions (temperature).    (Figure 9) of all samples is small (< 2 %) and negative and its value slowly increases with cooling. Negative magnetoresistance is rare in non-magnetic materials and usually has an exotic origin indicating again that SnO2 samples do not follow simple metallic behavior. However, negative magnetoresistance is found in impurity conduction of many semiconductors [25][26][27] and is often attributed to specific mechanisms of scattering on disorder which is in accordance with the conclusions extracted from the temperature dependence of DC resistivity and Hall coefficient. Figure 9 Magnetoresistance for SnO2 single-layer and bi-layer samples. Figure 10 Electron mobility for SnO2 single-layer and bi-layer samples. Figure 10 shows the temperature dependence of mobility μ calculated as μ = RH/ρ for all samples. As expected, like the DC resistivity and RH, the mobility is also temperature-independent, in contrast with simple metallic behavior where mobility and consequently resistivity decreases with decreasing temperature due to electron-phonon scattering.
There are several possible scattering mechanisms that limits transport properties in SnO2 thin films: scattering on phonons, scattering on ionized impurities provided by intrinsic and extrinsic doping and scattering on neutral defects (crystal imperfections, neutral impurities, grain boundaries …). According to theory only scattering on neutral defects is temperature independent and therefore presents the dominant scattering mechanism in our samples. The two single-layer (undoped) samples (S-590 and S-610) have practically the same mobility of about 10 cm 2 /V, which means that small variation in deposition temperature does not influence mobility. On the other hand, the two bi-layer (doped) samples (B-590 and B-610) show mobility difference of about factor 3. The B-590 sample has even slightly smaller mobility than the undoped samples, but the B-610 sample has a significantly greater mobility. Bi-layer samples are doped with equal fluoride concentrations which would mean that the structural properties related to deposition temperature at least for doped samples, strongly affects the scattering of free carriers.
From Table 2 can be seen that bi-layer samples have larger textured coefficient for (200) orientation. Comparing mobility values to preferred orientation present in sample ( Figure 11) it can be observed that mobility increases with increase of (200) texture coefficient. The same relation of mobility to texture coefficient was observed in ref. [28] . The mobility of sample B-590 is smaller than expected and this confirms that crystal orientation is not the only parameter that influences mobility but is the dominant one in our case. Mobility dependence on nanocrystal size is not observed. As explained in ref. [28] scattering at nanocrystal boundaries affects mobility only if the nanocrystal size is close to the mean free path of the charge carriers. Estimated mean free path for this type of samples (< 5 nm) is considerably shorter than the nanocrystal sizes of the present films (tens to fifty nanometres). Therefore, Hall mobility is unlikely to be limited by the nanocrystal boundary scattering.

Conclusions
We have examined transport properties of SnO2 thin films deposited by APCVD. Two different type of samples were deposited: undoped SnO2 single-layer and doped/undoped SnO2 bi-layer at two different deposition temperatures. Small variation in deposition parameters results in samples with different structural (nanocrystal size and preferred orientation) and transport properties (resistivity and mobility). Results of impedance spectroscopy and Hall probe suggest that transport properties of prepared SnO2 thin film samples are dominated by scattering at neutral defects. Mobility dependence on preferred orientation presented in samples is observed.