(Color online) Al dopant density in Si as a function of depth measured by ECV. The Gaussian-shaped characteristic of the profile decay is mainly caused by the lateral doping inhomogeneities within the measuring area. Thus, according to Ref. 15 , the determined profiles have been assumed as a convolution of the real Al profile and a Gaussian function. d Al stands for the depth, at which the profile decreases abruptly.
SEM image of the cross section of an Al-doped p + region at the Si surface. The Al doping profile ends abruptly, clearly indicated by the sharp contrast at the interface between the Al-p + region and the Si bulk. Thickness inhomogeneities within the ECV measurement area (Ø = 3.55 mm ± 0.06 mm) lead to the Gaussian-shaped characteristics of the measured profiles (cf. Fig. 1 ).
(Color online) Measured saturation current densities j 0,Al of the test samples (squares) with surface-passivated (open) and non-surface-passivated (closed) Al-p + regions and calculated influence of Auger recombination (lines) using an empirical Al doping profile (inset), without taking defect recombination into account. The profile depth and the minority carrier surface recombination velocity S 0,min have been varied. The inset shows the simplified profile used for our simulations.
(Color online) Measured saturation current densities j 0,Al of the test samples (squares) and calculated influence of Auger and SRH recombination (lines) for varied effective defect densities Nt * ≡ NA σ e . The two curves for each Nt * value indicate calculations for passivated (S 0,min = 100 cm/s) and nonpassivated (S 0,min = 107 cm/s) Al-p + surfaces, respectively.
(Color online) Measured saturation current densities j 0,Al of the test samples (squares) and calculated influence of Auger and SRH recombination with Nt * = 1.7 cm−1 (named the SRH model) and the lifetime parameterization (τAltermatt model) proposed in Refs. 1 and 2 .
(Color online) Gaussian function vs standard deviation σ. The medium variation of the measured depth Δd = 0.75 μm has been determined from fits to the measured ECV profiles (cf. Fig. 1 ). The different depths have been accounted for in our simulations (inset), according to their Gaussian probability.
(Color online) Measured saturation current densities j 0,Al of the test samples (squares) and calculated influence of Auger and SRH recombination without (dash dotted) and with (dotted) taking into account Al-p + thickness inhomogeneities (i.) (cf. Fig. 6 ). For non-surface-passivated thin Al-p + regions, a lower shielding for the minority carriers in the Si bulk leads to increased recombination activities at the surface and thus to an increase of j 0,Al. Taking into account incomplete ionization (i.i.) of the Al acceptors without (line) and with (dashed) Al-p + thickness inhomogeneities, the effect of the latter behaves similar. The lower amount of free majority carriers in the Al-p + region due to i.i. leads to a significant overall increase of j 0,Al and, furthermore, to a more pronounced difference for passivated and nonpassivated surfaces.
(Color online) Effect of incomplete ionization dependent on the acceptor concentration NA . Shown are the results of our simulations with the integrated model in Sentaurus TCAD (line) and calculations by Huster and Schubert (Ref. 15 ).
(Color online) Measured saturation current densities j 0,Al of the test samples (squares) and calculated (squares) influence of Auger and SRH recombination taking into account incomplete ionization of the Al acceptors. In this case (compared to Fig. 7 ), each simulation result is calculated with the corresponding measured profile at the same Al profile depth, respectively. The influence of profile inhomogeneities has been neglected. Excellent agreement of measured and calculated j 0,Al values has been achieved.
Parameters used for our simulations of the saturation current densities j 0,A l of Al-doped Si.
Defect level and capture cross sections for Al-contaminated Si reported in literature.
Experimental data and simulation results for our n-type Si solar cell with Al-p + rear emitter. By counting the Al atoms simply as acceptor occupancies (1), calculations show an overestimation of the open-circuit voltage V oc of 2.5%. While incorporating the τAltermatt model (2) still leads to a significant overestimation, the model presented in this work (3) shows very good agreement with the measured values.
Article metrics loading...
Full text loading...