Oscillator Strengths for Lines of Ionized Uranium (U II)

Oscillator strengths for 49 lines of U II recently measured by Voight can be used to calibrate the intensity scale of the U II lines in the NBS Tables of Spectral-Line Intensities and derive a larger set of oscillator strengths of lower precision but consistent with the new measurements. The standard deviation of the differences between the two sets of gf-values for the 49 lines is 29 percent. Oscillator strengths of that precision are given for 776 additional lines from the NBS Intensity Tables. The uncertainty in absolute value is 67 percent.


Introduction
Recently Voigt [1975] 1 meas ured oscillator strengths for 49 lines of U II in a wall-s tabilized argon arc. Hi s analysis of errors leads to the conclusion that , on a relative scale, hi s errors do not e xceed 10 perce nt. In his table III of res ults for U II lines he compares his values with gfvalu es from Corliss and Bozman [1962]. The well-known e nergy depe nde nt error of their valu es is clearly exhibited in this table. Nearly all the old values for lines originating from le vels above 25 000 cm -I are larger than Voi gt's valu es, whil e for lines originating from le vels below 25000 c m -1 the old values are nearly all smaller. A new calibration of the level populations in the co pper arc of Meggers, Corliss, and Scribner [1975] is clearly required.

. Comparison of Intensities with Oscillator Strengths
We compare the inte nsities from Meggers, Corliss, and Scribner with the new oscillator strengths by the usual population plot of log fA :J/ gfv versus upper energy level. Th e standard deviation of the residuals from the least-squares line fitte d to the 49 points was 0.20 dex (±58%). Since Voigt's gfvalues have a relative error of only 10 percent, we may interpret the residuals as mostly error in th e intensities.
The errors in the inte nsiti es may be random , systematic or both . Random errors cannot be re moved but sys te matic errors can be removed if they can be specified. There see m to be two possibilities of syste mati c error in the intensities, i. e., as a fun c tion of wavelength or of inte nsity. The residuals from the plot 1 Yea rs in brac kets indi ca te th e (jt er ature refere nce at the end of thi s pape r.
were plotted against both quantities and th ere was correlation in each case. It is not s urpri sin g that if one quantity were to s how correlation the other would also, since atomic s pectra us ually s how a correlation betwee n wavele ngth and inte nsity. How e ve r, in this case the correlation of th e residuals with inte nsity was better th a n with wavele ngth . This correlation plot a nd its least-squ ares fitted lin e are s hown in fi gure 1. The lin e can be re prese nted by the equation The inte nsit y correlation s hown above implies either that th e intensity scale of Meggers, Corliss. and To derive from our intensities gfvalues consistent with Voigt's we should remove this systematic effect.
By subtracting R from log I we obtain a corrected intensity which has the systematic error removed.
Log I'A 3/ gfv is then recomputed using the corrected intensity and plotted in figure 2. The standard deviation of the residuals is now reduced to 0.-11 dex ( ± 29%), which in fact represents also the standard deviation of the differences between Voigt's gfvalues and gfvalues derived from the corrected intensity scale of U II lines in the Tables of Spectral-Line  Intensities. . The equation of the least-squares fitted line in figure 2 is log lA 3 /gf= 16.531-0.00006717 E. With this equation we calculate gffor the 49 lines measured by Voigt and for 776 others.

Results
The results are given in the tables. In table 1 we give log gf as measured by Voigt and as calculated from the Tables of Spectral-Line Intensities. The differences for the 49 lines are given in the fourth column. The standard deviation of the differences is 2 Be ll a nd Upson [1971] investigat ed the int ensity scale of Meggers , Corliss, a nd Scribne r for th e case of Fe I and co ncluded th at the Fe I scale was too compressed. That concl us ion does not support the pres ent res ult. Laboratory. Earlie r references for levels are given in the Intensity Tables. In a number of cases th e intensities in Meggers, Corliss, and Scribner re prese nt the summation of unreso lved pairs of lin es. When both of the lines origin a te from th e ion , we divided the inte n sity ac· cording to the ratio gi ve n in Steinhaus e t al. [1972].
The re mainin g unresolved pairs we re not use d in thi s paper.
Th e error in th e re lative scale of gf as de te rmined a bove is a bo ut 30 percent. T o calc ul ate th e e rror in the a bsolute scale, we add (quadrati cally) the abo solute error of 60 pe rce nt de termined by Voigt for his absolute sca le with whi c h we are calibra t e d. The un ce rtainty in our absolute scale is thu s about 67 percent. This large absolute error arises from th e uncertainty in the co ntinuou s background which had to be subtracted from a faint U I lin e durin g Voigt's meas ure ment of the relativ e inte ns ity of a U I and a U II lin e in hi s wall-stabilized arc. A direct measureme nt of a life tim e in U 11 would avoid thi s source of e rror.