HCl gettering of multicrystalline wafers

The impact of gettering by HCl etching was initially investigated on a low-impurity level. Gettering by HCl gas could be an attractive alternative to phosphorus gettering for wafer cells. Furthermore, the use of 1 Ωcm mc silicon wafers used in the photovoltaic industry enables the measurements of the lifetime and fast measurement methods are available. In general, the contamination concentration of e.g. iron in mc is between 10¹² and 10¹⁵ cm^-3. As the energy levels of transition metals are located in the band gap of silicon, recombination of carriers via these traps occurs. The contamination level determines the lifetime of a wafer. The most prominent defect in silicon is iron, as it has normally the highest impurity level and is nicely detectable. In the boron-doped silicon, interstitial iron is known to form FeB pairs, which are in turn dissociated by optical activation or thermal annealing. The lifetime increases after dissociation, as the FeB energy level is relatively shallow in contrast to the deep centre of the interstitial iron Fe_i [101]. Assuming that all

other recombination processes remain unchanged before and after the dissociation and that 100% of the iron is paired or interstitial before and after the dissociation, the total interstitial iron concentration can be calculated as follows:

( )

where τFei and τFeB are the lifetimes of interstitial iron and paired FeB depending on the injection level. The prefactor C depends on the dopant densities and injection levels [101]. The determination of the intersital iron concentration [Fe_i] is difficult for concentrations smaller than 1x10¹⁰ cm^-3. Besides the measurement error in lifetime, the illuminated condition is instable as the amount of iron atoms is much smaller than the amount of boron. However, for higher impurity amounts this method is a well-suited and fast detection technique.

The wafers used for the experiments were from the middle of the blocks and cut in different columns. Process variations were applied on vertical neighboring wafers to avoid influences due to the grain structure. Prior to the getter processes, the samples were etched by CP133 and cleaned with RCA.

Phosphorus diffusions were performed at 880°C and resulted in 16 Ω/sq. sheet resistances, which were taken as reference getter process. Furthermore, untreated wafers and corresponding wafers annealed at the etching temperatures and times were measured. Since HCl-etched samples sometimes exhibit pitted surfaces, all wafers were subsequently etched by CP133 removing approximately 3 µm. All samples then received a SiN_x passivation to suppress surface recombination processes. Lifetime measurements were performed by the quasi-steady state photoconductance (QSSPC) method [171] or by carrier density imaging (CDI) [172], allowing high-resolution lifetime maps. The iron concentration is then calculated from the QSSPC measurements at an injection level of 2x10¹⁵ cm^-3. The dissociation of the FeB pairs was accomplished by illumination. Due to the increased carrier mobility at 80°C, FeB pairs are formed after 5 minutes of annealing. The calculated Fei concentration therefore represents an area-average value, as the spot from the QSSPC measurements has an area of several cm².

6.5.2 Variation of temperature and HCl concentration

Preliminary experiments were performed on mc wafers from Scanwafer. HCl gettering was performed at temperatures of 1200°C or 850°C and 2% HCl

concentration for 30 minutes. Furthermore, a slow cooling ramp was examined in contrast to a fast cooling ramp. Cooling ramps have a strong influence on the mc quality, as the diffuse reflectance and solubility of impurities depends on temperature and causes e.g. supersaturation during cooling. This influences the final distribution and chemical state of the impurities in mc silicon [145].

It was clearly noticeable that the temperature of 1200°C dramatically decreases the lifetime of the sample despite of the HCl containing gas. It is assumed that most metal precipitates are interstitially dissolved at this high temperature, as the time was too short to extract all impurities. However, some lifetime improvements of the HCl-etched sample at 850°C could be perceived compared to the corresponding annealed sample. By using a weighted average lifetime⁸, a similar lifetime is then observed for the HCl-etched sample compared to the phosphorus gettered one. Surprisingly, no clear difference between the fast and slow cooling ramp was noticeable even though the slow cooling should allow the impurities to diffuse to the grain boundaries, where they are less detrimental. The subsequent experiments were carried out with the fast cooling ramp.

Experiments with mc wafers from Deutsche Solar were performed at a temperature of 850°C for 30 minutes and a variation of the HCl concentration from 2% to 30%, as the amount of available chlorine radicals at the surface could be a dominating factor. Figure 6-25 exemplarily shows the CDI measurement of column C depending on the different processes. These images are typical for mc material, as the lifetime depends on the grain structure and is lower at grain boundaries.

On the left-hand side in Figure 6-26, the lifetimes measured by QSSPC at an injection level of 2x10¹⁵ cm^-3 from wafers of a different column are shown. The overall high lifetime for the untreated wafer is unusual and implies a good temperature monitoring of the ingot crystallisation. The impurities may be present as metal silicide precipitates, which are accumulated at grain boundaries.

By annealing, the lifetime is dramatically reduced as the impurities are dissolved through the wafer. It is then shown that a phosphorus diffusion increases the lifetime of some grains, but does not reach the high lifetime of the untreated wafer.

8 Regions of low lifetime essentially limit the cell efficiency. Averaging the lifetime not by the arithmetic mean, but by the sum of the inverse square root lifetimes weights more these regions [173].

As cut Annealing P-diffusion

2% HCl 10% HCl 30% HCl

τ [µs]

Figure 6-25: CDI lifetime measurements of column C depending on the different processes.

The HCl getter processes were performed at 850°C for 30 minutes. The wafer size is approximately 6x5 cm².

Gettering with gaseous HCl shows an optimum HCl concentration for a maximum gain in lifetime which varies from 5% to 10%, depending on the column. By increasing the HCl concentration, spikes may be created inducing damage into the wafer and may hinder the diffusion of chlorine atoms towards the surface and of the metal silicides away from the surface. The total interstitial iron concentration (also shown in Figure 6-26) reveals no clear decrease with increasing HCl concentration within the measurement accuracy. As expected, the annealed wafers show the highest amount of interstitial iron. Compared to this, the iron concentration of HCl-gettered samples is decreased for at least one order of magnitude.

Because of the unusual high feedstock quality and a high deviation from one column to the others, the experiments were repeated on wafers of a different ingot. The results of one column are shown exemplarily on the right-hand side in Figure 6-26. The untreated material shows a low lifetime of less than 50 µs, which is decreased even further after annealing. Here, the optimum HCl

concentration is about 16%. Again, compared to the untreated and annealed samples, the iron concentration is reduced by more than one order of magnitude.

However, for this material the lifetime of the reference POCl3 gettered samples could not be reached with any HCl concentration. We conclude that there is a strong dependency on the feedstock material and the optimum HCl concentration depends on the crystal structure. A window between 5 and 16%

HCl concentration is found in which improvements due to the gettering are clearly visible.

Figure 6-26: Lifetime measured by QSSPC at 2x10¹⁵ cm^-3 and iron concentration for different HCl concentrations and two different mc substrate materials.

6.5.3 Dependence on time

As described above, the metals react with the chlorine radicals at the surface and are then transported away by the gas stream. Taking into account only this effect, a longer gettering time should decrease the amount of metals in the bulk.

Robinson et al. observed that increasing the time from 1h to 16h improved the lifetime dramatically. HCl oxidation with 1% HCl concentration for 1h at 1100°C preserved untreated lifetime of 10-40 µs, whereas 16h oxidation at 1200°C resulted in minority carrier lifetimes up to 300 µs [159]. The improvement by longer gettering should also be visible when etching with HCl gas in a hydrogen atmosphere.

Experiments were performed with 5% HCl concentration at 850°C on columns D and F of the high-quality feedstock. The gettering time was varied from 10 minutes to a maximum of 120 minutes, limited mainly due to practical reasons, as the process has to be supervised all the time. Figure 6-27 shows the results of the lifetime and iron concentration depending on the gettering or annealing time. It can be noticed that the longer the treatment, the lower the

lifetime. The annealed samples show a high amount of interstitial iron of more than 1x10¹² cm^-3. This is decreased by two orders of magnitudes for the phosphorus gettered wafer as well as for all wafers gettered by HCl gas for more than 30 minutes. Despite the fact that nearly all interstitial iron is present for the 10 minutes etched sample, this exhibits the highest lifetime of all HCl-gettered wafers, close to the lifetime reached by phosphorus gettering. It is assumed that due to the long thermal treatment, thermally induced stress and dislocations are created, which reduce the lifetime. For longer high-temperature treatments, precipitates containing metals with low diffusion coefficients are dissolved.

However, the time may be still too short for the metals to diffuse to the surface.

For example, titanium diffuses only 9 µm for a 30 minutes process at 1300°C.

After 12h, the titanium would diffuse to the surface of a 200 µm thick wafer, but assuming an etch rate of only 0.1 µm, about 70 µm silicon would be removed.

This is not a practicable process in the PV industry. Longer getter times with very low HCl concentrations may be a compromise between low silicon removal and the extraction of all impurities.

10⁹

Figure 6-27: Lifetime measured by QSSPC at 2x10¹⁵ cm^-3 and Fe concentration depending on the gettering time for column D (left) and F (right).

6.6