In order to check possible dependencies of Stark widths on temperature we have plotted the ratios $w_{\rm S}/N_{\rm e}$ versus temperature T for each one of the lines. Figure 5 shows an example of these plots for the 407.216 nm OII line. The available data in the literature are also included. Although in some cases the widths seem to show very slight tendency to decrease with increasing temperature, the scarcity of available data and the inherent uncertainties in the measurements make it difficult to establish any functional relationship, at least in the temperature range considered in this experiment. This behavior is common for every measured line.

In Table 2 the Stark FWHM value $w_{\rm m}$ is listed for $N_{\rm e}=10^{23}\,{\rm m}^{-3}$ obtained from the mean of the ratios $w_{\rm S}/N_{\rm e}$ . The electron temperature is assumed to be around 40000 K. The 32 measured OII lines, with their lower and upper transition levels assuming LS-coupling (Striganov & Sventistskii 1968), are arranged by transition arrays and multiplets according to upper level increasing energy. The number of data and mean is shown for each line, with the corresponding standard statistical error $\sigma$ . For comparison, Table 2 shows other experimental available data Platisa et al. (1975); Djenize et al. (1991, 1998). The accuracy of these data is specified for each case in the Table 2 caption and takes values around 15% according to the authors. The similar accuracy of the present work takes advantage of a much more extensive set of data and lines. In this way, more reliable uncertainties for each line can be derived from the resulting statistical dispersions. All data have been normalised to $N_{\rm e}=10^{23}\,{\rm m}^{-3}$ , assuming a linear dependence with $N_{\rm e}$ and dividing by this value. When available, the ratio of the measured values to theoretical values $w_{\rm m}/w$ is shown, i.e. semi-classical and semi-empirical approaches calculated in Griem (1974); Hey & Bryan (1977); Hey & Breger (1980); Dimitrijevic (1982) and also results of Puric et al. (1988). These are always for $N_{\rm e}=10^{23}\,{\rm m}^{-3}$ and taking, if possible, T=40000 K.

Our experimental results are usually in good agreement within the same transition arrays and multiplets, as can be seen in Table 2. Also there is good agreement with the theoretical data of Griem and with the measurements of Platisa et al. or Djenize et al. within 20% uncertainty ranges. Very reliable Stark data and good experimental profiles of OII emission lines are noticeably difficult to obtain. Also the different plasma experimental conditions, mainly the different temperature ranges, could explain some discrepancies like those corresponding to the 408.929 nm line. Agreement with Griem, Dimitrijevic and Hey & Breger is good. The work of Hey & Bryan perhaps slightly underestimates the widths, while the formulae given in the work of Puric et al. (1988) clearly tend to underestimate the widths when applied to $\rm 3s-3p$ and $\rm 3p-3d$ transitions of OII specie. Puric et al. (1988) calculated their formula within a semiempirical approach and studied width dependencies on upper level ionization potential working with results of species with different ionization stages, from OII to OV, but mainly OIII.

Taking into account an electron density uncertainty of 10% we can estimate the average final uncertainty of the Stark widths given in this work around 20%, although with a strong dependence on the line. Perhaps the main error source for the narrowest transitions (often the most intense) is the unknown kinetic emitter temperature in deconvolution procedure. For some other transitions the error comes from weakeness of plasma emission. In spite of the few reliable data which can be obtained for most of the difficult lines, particular uncertainties can be inferred from statistical dispersions. Future high accuracy measurements should extend and complete the lack of OII Stark experimental data.

**Table 2:** Experimental Stark widths (FWHM) $w_{\rm m}$ in pm of OII spectral lines and the ratios to different theoretical values $w_{\rm m}/w$ . All the values are normalised to $N_{\rm e}=10^{23}\,{\rm m}^{-3}$ . For each line we specify the number of measurements n taken in this work. [1] Platisa et al. (1975), measures in a pulsed arc, $N_{\rm e}=0.52 \ 10^{23}\,{\rm m}^{-3}$ , $T_{\rm e}=25900$ K, acc. 16%. [2] Djenize et al. (1991), measures in a pulsed arc, $N_{\rm e}=0.81 \ 10^{23}\,{\rm m}^{-3}$ , $T_{\rm e}=60000$ K, acc. 13%. [3] Djenize et al. (1998), measures in a pulsed arc, $N_{\rm e}=2.8 \ 10^{23}\,{\rm m}^{-3}$ , $T_{\rm e}=54000$ K, acc. 15%. [4] Griem (1974), theoretical treatment, $N_{\rm e}=10^{23}\,{\rm m}^{-3}$ , $T_{\rm e}=40000$ K. [5] Hey & Bryan (1977), based on Griem semiempirical formulae, $N_{\rm e}=10^{23}\,{\rm m}^{-3}$ , $T_{\rm e}=25900$ K. [6] Hey & Breger (1980), theoretical treatment, method I, $N_{\rm e}=10^{23}\,{\rm m}^{-3}$ , $T_{\rm e}=25900$ K. [7] Dimitrijevic (1982), semiclassical approximation, $N_{\rm e}=10^{23}\,{\rm m}^{-3}$ . Interpolated to $T_{\rm e}=40000$ K. [8] Puric et al. (1988), calculated dependences on the upper-level ionization potential, $N_{\rm e}=10^{23}\,{\rm m}^{-3}$ , $T_{\rm e}=42500$ K, acc. 30%
$\lambda$	Transit.	Multiplet	n	$w_{\rm m}$	$\sigma$	$w_{\rm m}$	$w_{\rm m}$	$w_{\rm m}$	$w_{\rm m}/w$	$w_{\rm m}/w$	$w_{\rm m}/w$	$w_{\rm m}/w$	$w_{\rm m}/w$
(nm)	array				(pm)	[1]	[2]	[3]	[4]	[5]	[6]	[7]	[8]
465.084	$\rm 3s-3p$	$\rm ^4P-^4D^o$	9	22.5	3.7	23.7			0.71			0.73	1.37
463.885	$\rm 3s-3p$	$\rm ^4P-^4D^o$	8	22.0	3.3			22.9	0.70			0.72	1.35
464.181	$\rm 3s-3p$	$\rm ^4P-^4D^o$	11	22.3	6.0			22.1	0.71			0.73	1.36
464.914	$\rm 3s-3p$	$\rm ^4P-^4D^o$	12	20.9	2.0	22.9			0.67	1.16	0.73	0.68	1.27
434.556	$\rm 3s-3p$	$\rm ^4P - ^4P^o$	6	25.8	4.4				1.19			0.96	1.77
436.689	$\rm 3s-3p$	$\rm ^4P - ^4P^o$	11	24.8	7.9	22.1			1.13			0.92	1.69
433.686	$\rm 3s-3p$	$\rm ^4P - ^4P^o$	7	20.2	5.6				0.93			0.75	1.39
431.714	$\rm 3s-3p$	$\rm ^4P - ^4P^o$	9	25.6	5.7	21.9			1.18	1.62	1.06	0.95	1.78
434.943	$\rm 3s-3p$	$\rm ^4P - ^4P^o$	9	25.3	9.0				1.16			0.94	1.73
431.963	$\rm 3s-3p$	$\rm ^4P - ^4P^o$	10	26.1	6.3				1.20			0.97	1.81
459.617	$\rm 3s'-3p'$	$\rm ^2D - ^2F^o$	8	27.2	4.1	24.6		15-13.6*		1.49	0.97
459.097	$\rm 3s'-3p'$	$\rm ^2D - ^2F^o$	11	22.3	7.6	25.4
435.127	$\rm 3s'-3p'$	$\rm ^2D - ^2D^o$	9	22.5	7.1
434.742	$\rm 3s'-3p'$	$\rm ^2D - ^2D^o$	5	31.1	8.3	21.9				1.87	1.21
408.512	$\rm 3p-3d$	$\rm ^4D^o - ^4F$	6	25.2	7.6				0.98				2.15
407.886	$\rm 3p-3d$	$\rm ^4D^o - ^4F$	9	20.0	4.0				0.78				1.71
409.294	$\rm 3p-3d$	$\rm ^4D^o - ^4F$	5	21.3	2.3	23.5			0.83	0.96	0.71		1.81
407.216	$\rm 3p-3d$	$\rm ^4D^o - ^4F$	11	23.9	2.0	22.7			0.92				2.04
407.587	$\rm 3p-3d$	$\rm ^4D^o - ^4F$	10	19.7	3.3	23.5			0.76				1.68
415.330	$\rm 3p-3d$	$\rm ^4P^o - ^4P$	7	29.3	4.9	25.6			1.13	1.19	0.93		2.38
415.654	$\rm 3p-3d$	$\rm ^4P^o - ^4P$	6	32.9	9.8				1.26				2.67
413.281	$\rm 3p-3d$	$\rm ^4P^o - ^4P$	9	29.9	8.3	25.4			1.15				2.45
412.148	$\rm 3p-3d$	$\rm ^4P^o - ^4P$	6	24.2	8.6				0.94				2.00
411.079	$\rm 3p-3d$	$\rm ^4P^o - ^4D$	7	21.9	5.4				0.81				1.81
411.922	$\rm 3p-3d$	$\rm ^4P^o - ^4D$	8	24.8	6.8				0.90				2.05
411.203	$\rm 3p-3d$	$\rm ^4P^o - ^2F$	5	34.3	8.0								2.83
444.821	$\rm 3p' - 3d$ '	$\rm ^2F^o - ^2F$	10	34.3	9.0
432.748	$\rm 3p'-3d'$	$\rm ^2D^o - ^2D$	6	38.2	5.1
430.382	$\rm 3d - 4f$	$\rm ^4P-^4D^o$	6	118.6	23.5		95.6
408.929	$\rm 3d - 4f$	$\rm ^4F - ^4G^o$	9	100.9	18.6		66.4
447.788	$\rm 3d - 4f$	$\rm ^2P - ^4G^o$	6	127.0	20.1
414.609	$\rm 3p'''-3d'''$	$\rm ^6P - ^6D^o$	7	43.2	10.0

* Djenize et al. (1998) take the 459.617 nm OII line as a doublet (459.600 and 459.617 nm). We found no evidence of this, neither in Striganov & Sventistskii tables (1968) nor in the DASNIST data base (1990).

4 Results and conclusions