Line Strengths and Lifetimes of Levels in Neutral Uranium

Relative intensities of 549 U I lines observed in a dc copper arc are used to derive transition probabilities and oscillator strengths. Upper limits to lifetimes for 65 levels in neutral uranium atoms are determined.


Introduction
Se paration of isotopes by excitation of excited s tates through ab sorption of laser radiation is a relatively new d evelopm e nt that has parti c ular appli cati on in separation of uranium iso topes. Knowledge of approximate values for oscillator stre ngth s of U ) lin es and life tim es of corres pondin g upper levels is use ful in tha t work. This paper re ports suc h valu es fo r 549 lin es and 65 le vels of U I.

Derivation of Oscillator Strengths
The observations from whic h the oscillator strength s re ported here are deri ved we re ta ke n from th e NBS Tables of Spectral-Line Inte ns iti es by Meggers, Corliss, and Scribne r [1961]. Those inte ns ities we re meas ured in a 10 A dc arc be twee n coppe r e lectrod es containing one atom of uranium for e very 1000 atoms of coppe r. In 1962 Corliss and Bozman published oscillator stre ngths for 326 lines of U I. Th ey used an arc te mpe rature of 5100 K to obtain a relative scale which was the n adjusted to an absolute scale by comparison with othe r ele me nts for whic h absolute oscillator stre ngths we re the n known .
In the 13 years that have elapsed since the n, much more accurate work on d etermination of oscillator stre ngths has been reported. The numb er of papers published on the subject now a mounts to nearly 2400, wh e reas in 1962 it was only 600. This greatly expanded and also improved knowledge of oscilla tor s trengths has given us a better understanding of the population di stribution of excited levels of atoms in the coppe r arc of Meggers , Corliss, and Scribn er. According to the study made by Corliss in 1962 , the bes t oscillator stre ngths then available indicated that the le vel population of atoms in our copper arc declin ed by a factor of 17 for every 10000 cm -\ rise in level valu e. Recent 1 measure me nts of oscillator stre ngth s show that the arc is in fact hotter , so that th e corresponding d ecline in le vel population s is more nearly a factor of 10. Thi s factor of 10 is used in the present work to calculate rela tive oscillator strength s in U I.
Sin ce about 1964 th ere have been many meas ureme nts of life tim es of atomi c levels by mean s of beamfoil spectroscopy and sin ce 1956, by th e me thod of delayed coin ciden ce. Th ese n ew meas ure ments mak e co nversion of relative scales to a bsolute scales mu ch m ore certain tha n pre viously. The method of delayed coin cid e nce has been used rece ntly by Klose [1975] to measure the life time of the U ) le vel at 27887 cm -1 a bove the ground sta te. I ha ve equated his valu e for this le vel, 7.3 ns , to th e reciprocal of the sum of th e tran sition probabilities of the two principal tran sition s (at 3584.88 and 4620. 23 A) from th e le vel to adjust th e rela ti ve values to the ab solute scale.

Accuracy of Results
The standard deviation of an individual de termination of the inten sity from Megger s , Corliss and Scribn er is about 32 percent. Klose's un certainty in the lifetime is about 15 perce nt. These two uncertainties affect all of the normalized transition probabilities equally. The log of the uncertainty in the value adopted for the rate of decline of population of levels of U I in the arc plasma is estimated to be ± 0.1 for every 10 000 cm -1 that the upper level departs from the level of normalization at 27887 cm -1. A line with an upper le vel separated about 5000 c m -I , for example , from that le vel would have an un certainty in th e log gf of ± 0.05 ari sing from th e population un certainty. The corres ponding uncertaintie s of 0.12 in th e intensity and 0.06 in th e life time whe n combined quadratically with th e c hosen example of uncertainty in population brings th e total error to 0.14 (an error of 38 percent in this case. More than half of the U I lines reported here arise from levels that lie within 5000 em -I of the normalized level.
Recently Voigt [1975] has measured oscillator strengths for 22 lines of U I. His results are systematically different from the values reported ohere. His values are smaller than mine below 4240 A and larger at longer wavelengths. It can also be said that his values are smaller than mine for upper energy levels above 24 000 emt and larger at lower energy levels. Whether the wavelength or the energy level is significant for the discrepancy is not clear.
The uncertainties in the level lifetimes calculated from the transition probabilities can be determined from the uncertainties of these transition probabilities but there remains an additional error arising from unobserved transitions from each level not included in the summation. The existence of this error requires that the lifetimes be regarded as upper limits. If, however, calculation of lifetimes is restricted to only those levels involving relatively strong lines, the error due to faint lines will be small. In this paper we have restricted the calculation of lifetimes to upper levels for which the sum of the intensities of downward transitions given in Meggers, Corliss, and Scribner is 20 or larger on their scale. Since the faintest observed lines are between 0.3 and 7 on that scale, depending on the spectral region, there is some assurance that the given upper limit will not be far above the correct lifetime.
As an example, we show below the lifetimes for three levels from unpublished measurements obtained by a method of laser excitation and subsequent isotope separation by Janes, Itzkan , Pike, Levy, and Levin [1975]  This limited comparison tends to support our reasoning.    Table 2 lists the upper levels in cm -I , the J-value of the le vel, the statistical weight (2 J + 1), the upper limit of the lifetime in nanoseconds 00-9 s), and the number of transitions in the summation. All of these levels are of even parity.