In-Situ Monitoring of Atmospheric Pressure Tin Oxide CVD using Coherent anti-Stokes Raman Scattering

 

M J Davis and M E Pemble

 

Department of Chemistry

University of Salford

Salford M5 4WT, UK

Abstract

Coherent anti-Stokes Raman scattering (CARS) has been used to investigate the mechanisms involved in the atmospheric pressure chemical vapour deposition (CVD) of tin oxide films on glass using dimethyltindichloride (DMT) and oxygen and/or water at 645°C. A novel compact CARS experimental system which operates using a collinear sampling geometry was designed and constructed around a conventional Nd:YAG/dye laser combination. This system facilitates the sampling of gas phase species at specific points within the reactor. This high spatial resolution has been used to investigate the variation in concentration of a major reaction product, methane as a function of position within the reactor. In addition, we report on the likelihood of detection of methyl radicals, which are believed to form via the decomposition of DMT, under growth conditions and the use of CARS in probing the gas-phase temperature within the reactor in an entirely non-invasive manner via analysis of the vibration-rotation structure of the CARS spectra associated with the N2 process gas.

Introduction

Tin oxide is a frequently used coating on a number of materials including glass. It can be deposited on glass by an atmospheric pressure CVD process using dimethyltindichloride (DMT) as a precursor with oxygen as the oxidant. This reaction has previously been studied by a number of workers [1,2] and attempts at modelling the process have been made [3]. The model and experiment have generally been compared on the basis of the concentrations of products in the exhaust of the reactor and the thickness profiles of the deposited films [3] or average concentrations throughout the reactor determined by spectroscopy [2]. However, this limited access to experimental data is not entirely satisfactory since the complex models necessary for a complete description contain many parameters.

There would be considerable advantage in access to species concentrations at many points within the reactor. If the gas dynamics are not to be disturbed in making the measurements a non-invasive technique is required. Spectroscopic methods are therefore well suited and several possible techniques could be envisioned. FTIR can be used, but this requires windows to be fitted to the reactor which not only presents the danger of disturbing gas flows but also renders movement of the sampling point difficult. Absorption techniques which use wavelengths to which the reactor walls are transparent, such as near-infrared, allow sampling anywhere, but have the disadvantage of giving an average concentration along the path of the beam. Conventional Raman spectroscopy lacks sensitivity and a perpendicular collection geometry, necessary to sample from a single point, is difficult in the restricted reaction space of many CVD reactors. Coherent anti-Stokes Raman scattering (CARS) provides good sensitivity from a very small region within the reactor, while access is simplified by the coherent nature of the process which produces scattering in a tight, laser-like beam. While CARS is a well established technique in combustion research [4], studies on CVD reactors are less common [5]. The aim of the work presented here was to demonstrate that CARS could provide data on variations of species concentrations within our elevated-temperature, laminar-flow, cold-wall atmospheric-pressure CVD reactor.

Experimental

The tin oxide coating reactor was of the cold-wall, horizontal laminar-flow type. It consisted of a cylindrical quartz tube of 105 mm diameter with the bottom half occupied by a graphite susceptor, heated using electric cartridge heaters and supporting the glass substrate. A quartz plate was suspended above the susceptor and substrate to create an 8 mm high reaction space with a length of about 280 mm. The DMT precursor was held in a stainless steel bubbler which was heated to 150 °C and through which nitrogen was passed. This gas stream was diluted with further nitrogen, mixed with oxygen then fed to the reactor. All gas lines were heated to 180 °C to prevent condensation of DMT. Gas flows were metered using tubular flow meters fitted with needle valves. Low concentrations of methane for calibration purposes were produced from pure methane using mass flow controllers.

The CARS set-up was built around a Spectron SL800 Nd:YAG laser and dye laser. The Nd:YAG laser had an internal frequency doubler resulting in 532 nm output with a maximum energy of about 300 mJ and a pulse length of approximately 10 ns. This was used to pump the dye laser which consisted of a transversely pumped oscillator and two longitudinally pumped amplifiers. The last mirror in the pump beam chain was only 30 % reflective, allowing some of the pump beam to exit from the dye laser to be used as the CARS pump beam. Typical output energies used were 10 mJ for the pump beam and 3 mJ for the Stokes beam. The rest of the optical system for combination of the beams and detection of the CARS signal was contained in a single box, producing a compact and rugged system which is shown schematically in Fig.1. The pump and stokes beams are made collinear using a dichroic filter which transmits the pump beam but reflects the Stokes beam.

The collinear beams exited the CARS box and passed to the beam positioning system around the reactor. This allowed independent movement of the sampling position horizontally along the reactor and vertically away from the substrate. The beams were focused into the centre of the reactor using an achromat lens with a focal length of 150 mm. This resulted in a region from which the CARS signal originated calculated to be approximately cylindrical with a diameter of 13 mm and a length of 1.2 mm. Even allowing for the fact that the focusing might not have been perfectly diffraction limited, it is fairly certain that over 95% of the CARS signal originated from a region within about 25 mm diameter and about 2 mm long.

The beams exiting the reactor were directed back to the detector side of the CARS box where dichroic filters reflected most of the pump and Stokes energy into a beam dump. The CARS signal was then passed through a series of filters to remove any remaining pump and stokes energy and was detected with a photomultiplier tube. The output from the PMT was fed to a boxcar integrator and averager which used a gate of about 30 ns. The integrated signal was digitised and recorded on a PC. In a typical experiment, the laser would be triggered at 10 Hz and ten pulses averaged for each point while the dye laser was scanned at 0.05 nm.s-1. This gave a point spacing of 0.05 nm but a sampling interval of 0.005 nm, corresponding to about 1.3 and 0.13 cm-1 respectively. The Raman shift scale was calibrated by recording the CARS spectrum of methane and setting the peak position to 2916 cm-1 [6]. All spectra are reported here as integrated output from the PMT and are uncorrected for instrument response

Results

Fig. 2 shows CARS spectra of DMT and methane in the reactor at 300 °C, a sufficiently low temperature for no decomposition of DMT to occur. The blank spectrum shown was obtained with just nitrogen flowing through the reactor and shows the real signal obtained due to the non-resonant scattering from the carrier gas. Methane shown a sharp peak, set by the wavelength calibration to 2916 cm-1. It should be noted that the slightly dispersive shape of the peak, with a portion lower than the background signal, is characteristic for a CARS resonance in the presence of a significant background and arises from interference between the coherent resonant and non-resonant components [6]. DMT gives rise to a broad signal with a maximum at about 2910 cm-1.

Fig. 3 shows spectra obtained at three different heights above the substrate at a position about one third of the way along the reactor. The spectra obtained with just nitrogen (blank) and methane (0.2 %) are shown at each position for reference. With just DMT in the reactor the broad band previously observed for DMT is seen with no clear sign of methane, indicating that little DMT has decomposed. However, with oxygen as well as DMT a significant methane peak is seen in the spectra at 1 mm with a substantial reduction in the DMT peak. Further away from the substrate less methane and loss of DMT are apparent and at 4 mm from the substrate the spectra with and without oxygen are fairly similar.

A similar series of spectra is shown for a position near the end of the reactor in Fig. 4. Even with only DMT in the reactor, some methane is clearly observed 1 mm from the substrate indicating that a significant amount of DMT has been able to decompose by this stage. With the addition of oxygen, methane is observed at all distances from the substrate. Interestingly, the spectral region at lower shift than the methane peak appears fairly similar in all cases, suggesting that the DMT concentration at this point along the reactor is little affected by the presence or absence of oxygen. However, it should be borne in mind that other species could give rise to intensity in these regions.

Spectra obtained at a number of points down the reactor at a height of 2 mm above the substrate are shown in Fig. 5. With only DMT in the reactor there is no clear evidence of methane until well down the reactor at 200 mm. However, with oxygen present, methane is apparent near the front of the reactor at 92 mm and is very abundant by the end. Again the results for DMT are less clear since other species may be contributing. However, the intensity in the region at slightly greater shift than methane does show a consistent decrease down the reactor. It is also worth noting that the spectra at 200 mm were recorded independently of those in Fig. 4 at 2 mm above the substrate (i.e. at the same point in the reactor) but show a qualitatively similar picture, including similar intensities for DMT with and without oxygen in the region at lower shift than the methane peak.

Discussion

The results presented are in qualitative agreement with the variations in concentration of methane and DMT which would be expected for this type of reactor. Since tin oxide is coated on the glass plate all the way down the reactor, reaction must also be occurring continuously down the reactor. Therefore the methane concentration should increase with distance from the inlet and DMT concentration should decrease, which is indeed observed. Since only the substrate is heated the lower part of the reactor will be much hotter than the upper parts, promoting a grater reaction rate here and products such as methane may be partly produced by surface reactions. Therefore methane concentrations would be expected to be higher nearer the substrate, particularly near the front of the reactor, which corresponds to the observed results. However, further down the reactor products from reaction near the inlet will have had time to diffuse away from the substrate, resulting in a lesser verticle product gradient here. Again, the results are in concordance with this reasoning. Therefore this indicates that CARS is indeed a suitable tool with which to obtain information on species concentrations from specific points within the reactor.

In order to provide an experimental comparison with model results it will be necessary to obtain quantitative results for a number of species over a complete grid within the reactor. Although the results shown here are only for precursor/products, the observation of intermediates would also be very helpful in improving models of the process.

One such intermediate which would be expected to form in this CVD process is the methyl radical, CH3•, and it is useful to consider if this could be detected in this reactor using CARS. This species has been detected using CARS, though in this case it was produced by photodissociation and its transient presence detected [7]. In our case, an essentially steady state population would need to be detected and, unfortunately, the likely concentration of methyl in the reactor is unknown. However, it would be expected to be several orders of magnitude, say 1000 times, less abundant than methane, one of the final products derived from it. The highest concentration of methane observed in the reactor during reaction was about 1 % and the detection limit can be estimated to be about 0.05 %. Even if the scattering cross section of methyl is assumed to be similar to that of methane the detection limit will still be 50 times higher than necessary for its detection. Reduction of signal noise in the spectra by compensating for energy and mode fluctuations in the lasers could be expected to reduce the detection limit by a factor of about ten, but this would still not be sufficient. Increasing the laser energy will not improve sensitivity due to the high level of non-resonant background signal produced in an atmospheric pressure reactor. Therefore, background suppression methods, which can provide a reduction in non-resonant signal of 10 times or more [8], would probably be necessary in order to observe intermediates such as the methyl radical.

An added complication to the modelling of cold-wall CVD reactors is that the temperature varies very considerably from point to point within the reactor and will have a strong bearing on the reactions occurring at any point. It would therefore be helpful to experimentally measure the temperatures at the sampling points in the reactor. This is possible with the CARS technique by observation of the vibrational-rotational band structure of an appropriate species. Nitrogen is eminently suitable since this has been frequently used for CARS thermometry and the relevant molecular parameters required for extraction of the temperature from the observed band are well known [9]. Again, this technique is widely used in combustion research were temperatures of around 2000 K can be measured to within 40 K [9]. For the lower temperatures generally present in CVD reactors temperatures should be measurable to within about 20 K. In our reactor the quartz top plate is not heated and there may be a gradient of as much as 450 K across the reaction space which, if assumed linear, is 56 K.mm-1. Thus even a laser based absorption technique using a beam with a diameter of about 2 mm will be sampling a reaction space where temperatures vary by 100 K or more. Across the CARS sample diameter of about 25 mm the temperature variation is only of the order of 2 K.

Conclusions

We have demonstrated that CARS is a good technique for highly spatially selective investigation of precursor and product species distributions within a laminar-flow cold-wall tin oxide coater. Estimates suggest that detection of intermediate species at much lower concentrations would require suppression of the non-resonant background in atmospheric pressure reactors. Given the significant temperature variations within the reactor, CARS should provide a very useful, non-invasive technique for measuring temperature profiles.

References

1.    D.A. Strickler, Ph.D. Thesis, 1989

2.    H. Sanders, Proceedings of the 14th International CVD conference and EuroCVD-11, 81-88 (1997).

3.    C.J. Giunta, D.A. Strickler, R.G. Gordon, J. Phys. Chem. 97(10), 2275-2283 (1993).

4.    D.A. Greenhalgh in Advances in Non-linear Spectroscopy (R.J.H. Clark and R.E. Hester, Eds.), Wiley, New York, 1988, pp.193-251.

5.    N. Herlin, M. Lefebvre, M. Péalat, J. Perrin, J. Phys. Chem., 1992, 96, 7063-7072.

6.    J.W. Nibler, G.V. Knighten in Raman Spectroscopy of Gases and Liquids (A. Weber, Ed.), Springer-Verlag, Berlin, 1979, pp. 253-299.

7.    B.K. Andrews, K.A. Burton, R.B. Weisman, J. Chem. Phys., 1992, 96(2), 1111-1120.

8.      A. Pott, T. Doerk, J. Uhlenbusch, J. Ehlbeck, J. Hoschele, J. Steinwandel, J. Phys. D - Appl. Phys., 1998, 31(19), 2485-2498.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 1.  Schematic of the CARS system. The beam paths are traced by the coloured lines as follows: green, pump; red, Stokes; yellow, collinear pump and Stokes; light blue, collinear pump, Stokes and CARS signal; dark blue, CARS signal. BD is a beam dump; D are dichroic filters as follows: D1 transmits 532 nm and reflects other wavelengths, D2 reflects wavelengths >580 nm, D3 reflects 532 nm only; H1 is a holographic notch filter for 532 nm; F are filters as follows: F1 is a 460 nm interference filter, F2 and F3 is a blue colour glass filters (BG12 and BG3 respectively).

 

Figure 2. CARS spectra of DMT and methane (0.1 %) in the reactor at 300 °C. The DMT bubbler was held at 150 °C, the bubbler flow rate was 0.6 SLM and the total flow rate was 8 SLM. The spectrum obtained with just nitrogen in the reactor (blank) is shown for reference.

Figure 3.  CARS spectra as a function of height above glass substrate for the DMT reaction with oxygen at a position 92 mm from the inlet. The substrate temperature was 645 °C and the total flow rate 4 SLM. The DMT bubbler was held at 150 °C with a carrier gas flow of 0.3 SLM. Spectra for two oxygen flow rates (a) 0.3 SLM and (b) 0.6 SLM are given. Spectra obtained with just nitrogen (blank) and also with 0.2 % methane are shown for reference.

Figure 4. CARS spectra as a function of height above glass substrate for the DMT reaction with oxygen at a position 200 mm from the inlet. The substrate temperature was 645 °C and the total flow rate 4 SLM. The DMT bubbler was held at 150 °C with a carrier gas flow of 0.3 SLM. Spectra for two oxygen flow rates (a) 0.3 SLM and (b) 0.6 SLM are given. Spectra obtained with just nitrogen (blank) and also with 0.2 % methane are shown for reference.



 

Figure 5. CARS spectra as a function of distance along the reactor for the DMT reaction with oxygen at a height of 2 mm above the substrate. The substrate temperature was 645 °C and the total flow rate 4 SLM. The DMT bubbler was held at 150 °C with a carrier gas flow of 0.3 SLM. The oxygen flow rate was 0.6 SLM. Spectra obtained with just nitrogen (blank) and also with 0.2 % methane are shown for reference.