Laboratory solar cells

An important characteristic of an RTP process is its thermal budget. In general the thermal budget is defined as the integral of process temperature over process time [23]. For the RTP processes developed in this work, the thermal budget reduces essentially to the product of plateau time and plateau temperature since heating and cooling rates of 100 K/s are applied.

A priori it is not obvious which thermal budget is best with respect to solar cell efficiency. In this section we investigate whether high-efficiency Cz-Si solar cells can be manufactured by very short RTP processes and how far the process time can be reduced without sacrificing solar cell efficiency.

Experimental

PV-grade Cz-Si from Deutsche Solar GmbH with a base resistivity of 0.9 cm was used. The wafers were 290 m thick and cut in pieces of 5 5 cm² in size. The saw damage was etched off in a CP133 solution and the wafers were submitted to a modified RCA clean. Then, 2 m of Al were evaporated on the back side and the P dopant P509 was spun on front. Subsequently, co-formation of emitter and Al-BSF was performed by means of RTP. The plateau time and temperature were varied, i.e. different thermal budgets were used. After removal of the PSG, the emitter surface was passivated by growth of a 10 nm thin RTO oxide at 950 C for 30 s). As the P509 yields considerable amount of inactive P, decreased during RTO. The initial RTD conditions were chosen to yield of 80 to 90 /sq after RTO. Front contacts were prepared by photolithography, evaporation of a Ti/Pd/Ag stack and Ag plating. FGA at 400 C for 30 min was applied, followed by deposition of a double layer TiO₂/MgF antireflection coating. Finally the cell edges were isolated by laser cutting and cleaving.

Solar cell results and analysis

Table 4.3 shows an overview of the employed thermal budgets and the resulting solar cell efficiencies for the cells fabricated with varying thermal budgets for RTP P-Al co-diffusion.

Apparently, by varying the diffusion time from 120 s down to 5 s, the best solar cell results have been achieved with a decreasing diffusion time and an increasing diffusion temperature. This is mainly due to an increasing" ", which we attribute to an improved emitter homogeneity for the shorter plateau times. It is well known, that during RTP the wafer edges are cooler than the wafer centre as they emit more radiation than they absorb. For the 5 s flash diffusion process a solar cell efficiency of 17.5 % has been achieved including 30 s of oxidation. To our knowledge this is the fastest ever diffused and oxidized silicon solar cell of such high efficiency. Including heating and cooling times, the overall process time for diffusion and oxidation was below 1 minute. This demonstrates that the fabrication of high efficiency Cz-Si solar cells does not necessarily imply lengthy gettering or annealing steps.

Tab. 4.3: Cz-Si solar cell results as a function of RTD plateau temperature and time. Each combination results in the same emitter sheet resistance after RTO (cell size is 4.6 4.6 cm ). Results before degradation.

" " ( [ C] [s] [mV] [mAcm^- ] [%] [%]

850 120 619 35.1 77.7 16.9

890 30 623 35.7 77.8 17.3

910 15 623 35.5 78.8 17.4

930 5 623 35.5 79.2 17.5

In order to assess the electron diffusion length# in the bulk of the 17.5 % efficiency solar cell, we first determined the lower and upper limit of the effective back surface recombination velocity by performing photo-current and voltage decay measurements (PCVD, see [5]) of solar cells made of 0.5 cm ( = 3.26 10¹⁶cm^-3) FZ-Si processed in just the same way as the Cz cells. On the basis of the high bulk lifetime of the FZ cells ( 400 s), an between 4000 and 7000 cm/s can be deduced from the measurements. Taking the linear relationship between and base doping according to Eq. 3.7, this transforms into 2000 to 3500 cm/s for the 0.9 cm Cz-Si ( = 1.7 10¹⁶cm^-3). By the application of spectrally-resolved light beam induced current measurements (SR-LBIC) the effective diffusion length # of the minority carriers can be measured. # takes into account the finite nature of a solar cell and thus is connected to # via a geometry factor which is a function of , the bulk thickness ) and the electron diffusion constant [4, 48]:

# #

!

! (4.1)

200 250 300 350 400 450 1000

10000

L

240 260

2000 3500 S

[c m /s ]

L

[µm]

Fig. 4.2: - plots for the maximum and the minimum obtained from SR-LBIC measurements of the 17.5 % efficient Cz-Si solar cell. For we can extract a lower value of 250 m and an upper value of 310 m.

10¹ 10² 10³ 10⁴ 10⁵ 10⁶

17,0 17,5 18,0 18,5 19,0

17.5 % RTP cell

250 310

E ffi ci en cy [% ]

S

[cm/s]

Fig. 4.3:PC1D simulation of the cell efficiency as a function of the effective back surface recombination velocity for the determined minimum and maximum bulk lifetime of the 17.5 % efficient RTP Cz-Si solar cell.

For the 17.5 % solar cell,# has been determined to be between 240 and 260 m. From these values it is possible to extract an upper and lower boundary for# from the corresponding

# - -plot in Fig. 4.2 This procedure leads to a lower value of 250 m and an upper value of 310 m for# , which corresponds to lifetimes of 22 and 37 s respectively. These are typical values for PV-grade Cz-Si indicating the preservation of the initial lifetime.

We used PC1D simulations to reveal weak points of the current cell design in order to gain knowledge of how to further increase the solar cell efficiency. The simulation has been carried out by taking into account all the information obtained by analyzing the 17.5 % solar cell, for example, quality of the second diode, series resistance, parallel resistance and reflectivity. In Fig. 4.3 the calculated solar cell efficiency is plotted versus for the deduced high and low values of # . Satisfyingly, the measured 17.5 % efficiency is well reproduced. But it becomes quite clear that the current solar cell efficiency is limited by the poor quality of the Al-BSF formed from the 2 m thin layer of Al evaporated prior to diffusion. Using thicker Al layers like screen printed Al would lead to deeper BSFs and thus to lower (see section 3.2). Alternatively, a boron BSF could be formed using spin-on dopants. Both methods would account for -values in the range of 500 cm/s and solar cell efficiency is predicted to increase up to 18.0 % for the PV-grade Cz-Si used here. As boron diffuses slower than phosphorus, an RTP step involving simultaneous diffusion might not be favorable. But, the current oxidation process could be modified to allow simultaneous formation of emitter passivation and single-sided boron diffusion. Boron spin-on dopants yielding the desired high boron concentrations are commercially available [34]. Of course, this would imply that the remaining boron glass has to be removed by a selective etching method, for example plasma etching (see e.g. [148]), in order to keep the passivation oxide on the front intact.