The spectra have been recorded by means of the Meudon Observatory Eagle mounting VUV 10-m spectrograph. The instrument is fitted with a concave (R= 10.685 m), 3600 lines/mm Jobin-Yvon holographic grating which gives a plate factor of 0.026 nm/mm in the first order. The slitwidth was 30 µm, giving a resolving limit of 8x10-4 nm throughout the whole range of observation (0.8 cm-1 at 100 nm). The presence of a magnetic field of about 0.1 tesla in the water cooled electric discharge lamp (ANVAR), made of a brass anode and two aluminum cathodes, makes it possible to run the source at a pressure as low as a few pascals, so that self-absorption is much reduced at short wavelength. With hydrogen flowing in the source, the lamp was operated at 350V, 200mA. Exposure time on Kodak SWR plates varied from 5 min to 13 h, depending on the spectral range photographed. Reference lines (CuII, GeII, SiII and 0I, NI, ClI, Cl II as impurities) were obtained by flowing helium in an ancillary water cooled copper hollow cathode discharge lamp operated at 500V, 200mA. Small pieces of germanium and silicon were introduced inside the cathode. Due to the difference of the grating illumination by the two lamps, one finds a random shift between the two spectra (Kaufman and Edlén 1974, Freeman and King 1977). For each plate the shift is determined by means of the impurity lines common to the spectra emitted by both lamps. In the shortest wavelength range (78-89 nm), the hollow cathode spectrum was recorded in the same zone of the photographic plate as the H2 spectrum, and extra Ar I reference lines were also obtained by flowing argon in the low pressure discharge lamp itself. The plates were measured on the Meudon Observatory photoelectric comparator giving an accuracy of 1mm on line position (Launay 1975). The measuring machine is now connected to a PC compatible micro-computer. The standard deviation of our polynomial fit for the hollow cathode reference lines is typically 7xl0^-5 nm and the wavelength accuracy of the hydrogen unblended lines is estimated to be±15xl0^-5 nm.

Relative intensity measurements have been performed by replacing the plate-holder of the spectrograph by a scanning photoelectric device moving along the Rowland circle and including a slit and either a photomultiplier with a MgF₂ window (Hamamatsu R1459) or a channeltron (RTC X919BL), depending whether the measured range is above or below 120 nm. The width of entrance and exit slits was 30mm. A set of 29 scans was obtained, each section of the spectrum being scanned at least twice. The relative intensities derived from raw data were scaled in order to get a consistent set of values over the whole spectral range (100 nm) but the instrumental response was not taken into account.