4.4 Block-cast multicrystalline
4.4.3 Bulk lifetime response to RTD P diffusion gettering
The solar cell results presented in the previous chapter indicated a decreasing bulk carrier lifetime with increasing RTD temperature. However, the used cell structure was not sensitive enough to fully reflect this behavior. Furthermore, the RTD induced change of the as-grown wafer could not be assessed. For this reason bulk lifetime measurements were performed on neighboring wafers coming from exactly the same middle part of the ingot as the ones used for the solar cells.
Experimental
The wafers were damage etched, cleaned and diffused exactly like those processed to solar cells.
The diffusion temperatures ranged from 860 to 1000 C and the diffusion time was adjusted to obtain similar of about 100 /sq. From 860 to 950 C, we used the Filmtronics spin-on dopant P509 and from 950 to 1000 C the less concentrated P507 source. The back surface remained uncoated. Heating and cooling rates were set uniformly to 100 K/s. In order to get access to the as-grown lifetime, one of the neighboring wafers underwent all of the above and following surface treatments apart from RTD. An overview of which process was applied to which wafer is given in Tab. 4.6. After RTD, the PSG and the emitter were etched off and both surfaces were passivated by deposition of a PECVD-SiNx. We cannot exclude that some hydrogen passivation of the bulk took place during the PECVD surface passivation but we assume this effect to be negligible. The SiNx layer yields surface recombination velocities in the range of only 10 cm/s [105] which is low enough to interpret the actually measured effective lifetimes as bulk lifetimes.
The lifetime was measured spatially resolved using the carrier density imaging technique (CDI) recently developed at the Fraunhofer ISE. Details on the method can be found in Riepe et al. [138] and Isenberg et al. [71]. CDI is based on the free-carrier absorption of IR-light in silicon. A hotplate (black body) emits IR radiation, which is transmitted through the silicon sample under investigation. A fast CCD-camera sensitive in the mid-infrared (3.5 to 5 m wavelength) measures the IR-transmission of the sample in two different states: In the first half of the lock-in period the sample is illuminated by a semiconductor laser ( = 917 nm) that generates an excess free-carrier density which is approximately equivalent to the generation at 1 sun. In the second half, the sample is in complete darkness, i.e. without excess carrier generation. The difference between the two images is proportional to the IR absorption of the excess free-carriers and thus to the local excess free-carrier density. Since the generation rate is known, actual lifetimes may be calculated. In contrast to the MW-PCD technique the CDI method measures actual lifetimes instead of differential lifetimes.
Results and discussion
Fig. 4.9 comprises the lifetime images as measured with the CDI technique. The striking areas which will be discussed in the following are marked from W to Z. The distributions are shown in Fig. 4.10 and the calculated mean lifetimes are given in Tab. 4.6 and in Fig. 4.12. For the as-grown, the RTD a and the RTD d diffused samples, the lifetime scans following a virtual line are compared in Fig. 4.11.
Tab. 4.6: Overview of the RTD diffused wafers from the middle region of an ingot and the respective mean carrier lifetime as deduced from the CDI lifetime mappings.
RTD seq. [ C [s] P SOD Wafer# in ingot Mean [ s]
as grown 258 24.1
a 860 150 P509 259 35.2
b 890 50 P509 261 26.5
c 920 18 P509 263 20.9
d 950 6 P509 265 15.4
e 950 60 P507 267 18.0
f 1000 18 P507 269 16.0
The wafer representing the as-grown state exhibits a comparatively homogeneous distribution of almost Gaussian shape. The mean value is 24.1 s and very few spots exhibit
values below 5 s. Compared to the as-grown reference, the diffused samples exhibit much wider distributions. A close look at the mappings reveals that there are various wafer regions which respond differently to the diffusion step. Some degrade, some remain almost unchanged and some improve significantly.
For example, in the as-grown state the grain boundaries of this multicrystalline material are hardly visible in the mapping and show medium lifetimes. This means that, at least in the as-grown state, most of the grain boundaries do not represent places of increased recombination.
However, this situation changes drastically after the RTD high-temperature step. As can be seen nicely in the Z marked region for instance, the grain boundaries and their very close surrounding become visible after diffusion showing much lower values than before. This becomes also visible in the one dimensional scans of Fig. 4.11. Compared to the as-grown state, their recombination activity increases significantly upon RTD. Also, after RTD d, the grain boundaries become more pronounced and exhibit even lower lifetimes compared to RTD a. The actual grain structure visible after RTD shows almost no correlation with the initial variations in the same region in the as-grown state. This is best demonstrated by comparing the W marked region in the as-grown with the mapping after RTD. Clearly, lifetime variations in the as-grown state and the existence of grain boundaries need not correlate. The pronounced activation of grain boundaries during a high temperature step was also observed after conventional furnace diffusion (CFD) and hence is not specific of RTD [2]. There are two conceivable explanations for the activation. Metal contaminants (e.g. iron) might be present in a passive state at the grain boundaries already in the as-grown state and might become active during the high-temperature
X
Fig. 4.9: Carrier lifetime of surface passivated block-cast wafers before (as-grown) and after single-sided P diffusion using varying RTD temperature and time combinations: a) 860 C/150 s, b) 890 C/50 s, c) 920 C/18 s, d) 950 C/6 s, e) 950 C/60 s, and f) 1000 C/18 s. The sketched line scans are shown in Fig. 4.11. Mappings were obtained with the CDI technique.
0 10 20 30 40 50 60 70 80 90 100 0
500 1000 1500 2000
as grown 860°C, 150 s 890°C, 50 s 920°C, 18 s 950°C, 6 s 950°C, 60 s 1000°C, 18 s
C ou nt s
Lifetime [µs]
Fig. 4.10:Distribution of the spatially resolved carrier lifetime of neighboring block-cast wafers before and after single-sided P diffusion using different RTD conditions.
step. Alternatively the grain boundaries may be comparatively clean in the as-grown state but act as efficient internal gettering sites for impurities present in the surrounding few microns and hence become decorated during the diffusion step [160].
The X and the Y marked region in the mappings are representatives of areas with low initial lifetime where typically small grains and a high density of grain boundaries and supposedly dislocations prevail. In some of these areas, such as region X for instance, the lifetime decreases further after RTD. This decrease occurs already after RTD at 860 C (process a) and seems to be the more pronounced the higher the diffusion temperature2. Presumably, precipitates of metallic impurities dissolve predominantly during high-temperature steps, leading to a net increase in the concentration of recombination sites [80]. However, the lifetime of region Y hardly changes upon RTD at 860 C and decreases only slightly during RTD at higher temperatures. There must be a fundamental difference in the impurity and crystallographic defect spectrum between region X and Y.
In areas of rather large grains, which typically show medium or high initial lifetimes, the lifetimes increase further during RTD. Region Z is a representative of those areas. Apparently, the improvement is higher for lower diffusion temperatures and for longer diffusion time.
Consequently, the largest improvement is obtained with RTD a. This can be best seen in the linescan of Fig. 4.11. Between the distance 30 and 40 mm, it intersects region Z. Obviously, of the large grains exceeds 60 s after RTD a. In contrast, no improvement but degradation
2A similar behaviour can also been observed in the linescan of Fig.4.11 around the distance of 10 mm.
RTD a RTD d as-grown linescan
0 10 20 30 40 50
1 10 100
as-grown
RTD a (860 °C, 150 s) RTD d (950 °C, 6 s)
Li fe tim e [µ s]
Distance [mm]
Fig. 4.11: One dimensional lifetime scan along the lines depicted in Fig. 4.9 for the as-grown and the RTD a and RTD d diffused samples.
of the same grains occurs with RTD d featuring 6 s of diffusion at 950 C. Hence, elevated temperatures seem to be detrimental. However, according to the mappings after 60 s of RTD at the same temperature, a clear lifetime improvement within the large grains has occurred. And even process f) with 18 s of RTD at 1000 C is superior to RTD d with respect to improvement of the large grains. These observations are consistent with P diffusion gettering of Fe for instance because it fits to the mean distance the interstitial Fe atoms may diffuse during the RTD step [114, 74]. The possibility of being gettered by the P emitter is the higher the longer the mean diffusion path. The final Fe concentration is determined by the segregation coefficient which in turn decreases with increasing diffusion temperature and thus will limit the minimum achievable Fe concentration in the case of the high diffusion temperatures [22, 114].
Fig. 4.12 shows the calculated mean lifetime of the RTD diffused samples and, for comparison, those of the neighboring wafers from the same ingot which were P diffused in a conventional tube furnace for 15 min at different temperatures. The applied CFD process featured pulling of the wafers out of the furnace as quick as possible. No slow cooling ramps were applied. Please note that CFD implies the use of POCl3 which causes P diffusion and hence gettering from both surfaces. For diffusion temperatures exceeding approximately 900 C
860 890 920 950 980 1010 0
20 40 60 80
18 s 60 s
18 s 6 s 50 s t=150 s
as-grown RTP
CFD (15 min)
M ea n lif et im e [ µ s]
Diffusion temperature [°C]
Fig. 4.12: Mean (bulk) carrier lifetime of neighboring block-cast wafers after single-sided RTD P diffusion at different temperature/time combinations and after 15 min of conventional furnace diffusion (CFD) at different temperatures. CFD employed P diffusion from both side by the use of POCl and quick pulling of the samples out of the furnace. Dotted lines are to guide to the eye.
and the respectively short diffusion times the average lifetime of the RTD treated samples drops below the mean lifetime of the as-grown wafer. In case of CFD the mean lifetime also decreases with increasing diffusion time. However, at temperatures below 900 C the CFD process yields a considerably pronounced bulk improvement compared with the RTD process conducted at the same temperature. For the 15 min CFD this improvement vanishes only at 950 C. The results show that real RTD with diffusion times in the range of seconds is too fast and requires too high diffusion temperatures which does not comply with the desired lifetime improvement of block-cast multicrystalline Si in order to fabricate high-efficiency solar cells.
The results are in accordance with model calculations segregation induced gettering of impurities in Si, e.g. by P diffusion gettering (see e.g. [22, 114, 94]). Higher diffusion temperatures allow a faster diffusion of impurities and hence a faster gettering. However, this effect is partly offset by a reduction of the segregation coefficient, because the solubility of impurities in the bulk grows faster with the temperature than the one in the phosphorus layer.
Residual impurity concentrations are therefore lower at lower diffusion temperature, providing that the proper gettering time is chosen. To overcome this problem Plekhanov et al. [132]
suggested a variable temperature gettering process with several slow ramping down steps. This process allows to take advantage of the large gettering rate at high temperature for shortening of the total gettering time as well as of the low residual impurity concentration attainable at lower temperatures. However, the respective process times are still in the range of hours.
When processes of industrial significance are considered, i.e. processes featuring comparatively high heating and cooling rates, sufficient P diffusion gettering of lifetime-limiting metal impurities, such as Fe, requires diffusion temperatures below 900 C and diffusion times 15 min. For this purpose, CFD is definitely superior to RTD.