Usually the emitter of high-efficiency silicon solar cells has a low surface concentration of phosphorus, NS, and a silicon oxide layer for surface passivation.
The standard way to diffuse such an emitter is a two-step process. In the first step a shallow emitter with a high surface concentration is diffused. In the Fraunhofer ISE clean room this is usually done in a tube furnace. Nitrogen flows through liquid POCl3 and carries it to the quartz tube. A flux of pure oxygen is also led into the tube, oxidises the silicon and builds a phosphorus pentoxide layer. Furthermore chlorine compounds clean the surface when they are formed with metals which might be present on the wafer. The P2O5 layer created with this process serves as an unlimited source for phosphorus atoms (compare Fig. 5.1 left).
Fig. 5.1: Sketch of development of phosphorus doping concentration ND with increasing diffusion time for an unlimited doping source (calculations performed with PC1D [90]) and for a limited doping source (calculations according to Sze [91]).
The process temperatures are around 800 °C. After removal of the phosphorus-silicate-glass (PSG), which is formed during the diffusion, this shallow emitter serves as a limited doping source for the second step, the drive-in diffusion. This diffusion is performed at very high temperatures (about 1000 °C) and a gaussian profile, i.e. a deep emitter with low surface doping concentration, evolves. Since the drive-in diffusion in the standard high-efficiency process at ISE is performed in oxygen ambient, the front and rear surface simultaneously become very well passivated by growing a silicon oxide layer of about 105 nm thickness.
Unfortunately, the process temperature of 1050 °C is much too high for multicrystalline silicon (see chapter 3) and the standard process will have to be replaced by a new process sequence. A low surface concentration is desirable as surface recombination decreases with decreasing doping concentration [92]. This finding definitely holds for planar surfaces but not necessarily for textured samples [12,93]. Therefore different emitters were processed on samples with planar surfaces as well as on single-sided plasma-textured samples, a process which was developed within this work (see chapter 4). The material used was 1 Ω cm FZ silicon produced by Wacker Siltronic and 1.5 Ω cm multicrystalline silicon produced by ScanWafer.
In a first approach, shallow emitters were diffused for one hour at 800 °C and 820 °C. The subsequent oxidation at 1050 °C was replaced by a modified wet oxidation (details see section 6.2.1) for four hours at 800 °C, 850 °C and 900 °C.
An oxide layer for the rear surface in the range of 100-150 nm on the p-type bulk
material had proven to be well suited for the LFC process. To avoid further growth, the gas ambient was switched from oxygen to nitrogen during the process according to Table 5.1. Whereas a suitable oxide thickness between 100 and 150 nm was obtained on 1 Ω cm p-type base material, the oxide on the (still highly doped) emitter was about 200-300 nm thick. This was far too much for a good anti-reflection coating. Therefore, the oxide was removed in HF and the wafers were oxidised for 20 min at 900 °C with a dry oxidation process which resulted in oxides of about 10-15 nm. This was thin enough to allow for an adapted double layer antireflection coating in order to minimise reflectance. In the second approach, the drive-in oxidation was not performed and the thin oxide was grown directly after removal of the PSG.
Table 5.1: Process parameters for the wet oxidation processes used for drive-in diffusion.
The oxide thickness was measured on <100> orientated 1 Ω cm FZ silicon.
T
The emitter profiles of the different diffusions were measured by secondary ion mass spectroscopy (SIMS) by the company RTG, Berlin, Germany (Fig. 5.2).
After the first emitter diffusion at 800 °C, the total amount of donor atoms was the same in all wafers. The profiles in graph A (upper left) did not show the expected characteristics of an emitter with drive-in diffusion. According to theory the integrated amount of phosphorus atoms should still be the same after the drive-in but the distribution should be different, i.e. the deeper profiles should have a smaller surface concentration of donors NS after oxidation. But the deeper profiles also had a higher surface concentration and consequently a higher overall doping.
Ω Ω Ω Ω Ω
Ω
Ω Ω
Ω
Fig. 5.2: Emitter profiles measured by SIMS and sheet resistance measured with a four point probe. The emitters in the upper half were oxidised for four hours with a wet oxide. The emitter in the lower left was driven-in with a dry oxide for 1 hour. The thick oxides were removed and a thin oxide below 20 nm was grown for surface passivation. The lower right graph shows the profile of the shallow emitters which were not driven-in.
Nevertheless, the profile had a gaussian shape, indicating a limited source for the drive-in as expected. For the emitter diffusion at 820 °C in graph B, the profiles followed the same trends on a higher doping level. This was due to the higher amount of phosphorus which diffused during the first step. The standard dry oxidation process at 1050 °C is shown in picture C. It revealed a higher sum of doping atoms than the profiles after emitter diffusion at 800 °C although the first emitter diffusion took place at 790 °C (and therefore should provide a smaller amount of donors). In picture D the two profiles of emitters without a long drive-in diffusion are presented. They were only oxidised shortly in dry ambient to passivate the surface. The surface concentration of donor atoms was in the range of 1020 atoms/cm3. The diffusion at 825 °C was performed in a different furnace than all the other diffusions, thus the diffusion process parameters are not directly comparable.
The measured emitter sheet resistance of all profiles corresponded well with calculations performed using the mobility model of Masetti et al. [94], this adds to the reliability of the measurement and proves that no samples were mixed up. For an explanation of the unexpected doping profiles of the wet drive-in oxidation, the redistribution of doping atoms during oxidation has to be considered. The doping profile after oxidation depends on the oxidation rate, the relative dopant diffusion rates in the oxide and the silicon and thus on temperature, oxidation ambient etc..
The dopant segregation coefficient m (m = ratio of equilibrium concentration of dopant between silicon and oxide) for phosphorus is greater than unity (m > 1) and thus usually a pile-up of phosphorus occurs [95]. However, in the case of wet oxidation of highly doped layers some of the phosphorus was apparently built into the silicon oxide and could consequently not contribute to the doping profile. The reason for depletion of the phosphorus instead of pile-up can probably be found in the very fast oxidation rates.