3 Stark widths and shifts measurements

The measured profiles were of the Voigt type due to the convolutions of the Lorentzian Stark and Gaussian profiles caused by Doppler and instrumental broadening. For the electron density and temperature in our experiments, the Lorentzian fraction was dominant (over 87$\%$). van der Waals (Griem 1974) and resonance (Griem 1974) broadening were estimated to be smaller by more than one order of magnitude in comparison to Stark, Doppler and instrumental broadening. The standard deconvolution procedure (Davies & Vaughan 1963) was computed using the least squares algorithm. The Stark widths were measured with $\pm 12\%$ error. Great care was taken to minimize the influence of self-absorption on Stark width determination. The opacity was checked by measuring relative line intensity ratios within the multiplet No. 2 during the plasma decay. The obtained values were compared with calculated ratios of the products of the spontaneous emission probabilities (A) and the corresponding statistical weights (g) of the upper levels of the lines. The necessary atomic data were taken from Wiese et al. (1966), Lide (1994) and NIST (2000). It turns out that the experimental relative line intensity ratios are constant during the whole plasma decay period, testifying to the absence of self-absorption and agree, within $\pm 11\%$ accuracy, with calculated relative intensity ratio values.

The Stark shifts were measured using the well-known method based on the line center position observations during the whole plasma decay period (Djenize et al. 1998 and references therein). The Stark shift data were corrected for the electron temperature decay (Popovic et al. 1992). Stark shift data are determined with a $\pm$0.0008 nm error at a given N and T.