Before proceeding to readjustments of coupling constants so as to improve the agreement of the SO splittings with data, we analyze the influence of timeeven (mass and shape) and timeodd (spin) polarization effects on the neutron SO splittings. Based on results presented in the preceding Section, we calculate the SO splittings as
Figure 2 shows the SO splittings calculated using SLy4  the functional based on the original SLy4 [19] functional with spin fields readjusted to reproduce empirical Landau parameters according to the prescription given in Refs. [21,22]. Plotted values correspond to results presented in Tables 1 and 2. The results are labeled according to the following convention: open symbols mark results computed directly from the s.p. spectra in doublymagic nuclei (bare s.p. energies). These bare values contain no polarization effect. Gray symbols label the SO splittings involving polarization due to the timeeven mass and shapedriving effects, i.e., those obtained with all timeodd components in the functional set equal to zero. Black symbols illustrate fully selfconsistent results obtained for the complete SLy4 functional. Gray and black symbols are shifted slightly to the lefthand (righthand) side with respect to the doublymagic core in order to indicate the hole (particle) character of the SO partners. Mixed cases involving the particlehole SO partners are also shifted to the right.
The impact of polarization effects on the SO splittings is indeed very small, particularly for the cases where both SO partners are of particle or hole type. Indeed, for these cases, the effect only exceptionally exceeds 200keV, reflecting a cancellation of polarization effects exerted on the partners. The smallness of polarization effects hardly allows for any systematic trends to be pinned down. Nevertheless, the selfconsistent results show a weak but relatively clear tendency to slightly enlarge or diminish the splitting for hole or particle states, respectively.
The situation is clearer when the SO partners are of mixed particle ( ) and hole ( ) character. In these cases, the shape polarization tends to diminish the splitting quite systematically by about 400500keV. This behavior follows from naive deformed Nilsson model picture where the highest members of the ( ) multiplet slopes down (up) as a function of the oblate (prolate) deformation parameter. As discussed in the previous Section, the timeodd fields act in the opposite way, tending to slightly enlarge the gap. The net polarization effect does not seem to exceed about 300keV. In these cases, however, we deal with large orbitals having also quite large SO splittings of the order of 8MeV. Hence, the relative corrections due to polarization effects do not exceed about 4%, i.e., they are relatively small  much smaller than the effects of tensor terms discussed below and the discrepancy with data, which in Fig. 2 is indicated for the neutron 1f SO splitting in Ca. These results legitimate the direct use of bare s.p. spectra in magic cores for further studies of the SO splittings, which considerably facilitates the calculations.
As already mentioned, empirical s.p. energies are essentially deduced from differences between binding energies of doublymagic core and their odd neighbors. Different authors, however, use also oneparticle transfer data, apply phenomenological particlevibration corrections and/or treat slightly differently fragmented levels. Hence, published compilations of the s.p. energies, and in turn the SO splittings, differ slightly from one another depending on the assumed strategy. The typical uncertainties in the empirical SO splittings can be inferred from Table 3, which summarizes the available data on the SO splittings based on three recent s.p. level compilations published in Refs. [7,8,9].

Instead of largescale fit to the data (see, e.g., Ref. [16]), we propose a simple threestep method to adjust three coupling constants , , and . The entire idea of this procedure is based on the observation that the empirical SO splittings in Ca, Ni, and Ca form very distinct pattern, which cannot be reproduced by using solely the conventional SO interaction.
The readjustment is done in the following way. First, experimental data in the spinsaturated (SS) nucleus Ca are used in order to fit the isoscalar SO coupling constant . One should note that in this nucleus, the SO splitting depends only on , and not on (because of the spin saturation), nor on (because of the isospin invariance at ), nor on (because of both reasons above). Therefore, here one experimental number determines one particular coupling constants.
Second, once is fixed, the spinunsaturated (SUS) nucleus Ni is used to establish the isoscalar tensor coupling constant . Again here, because of the isospin invariance, the SO splitting is independent of either of the two isovector coupling constants, or . Finally, in the third step, Ca is used to adjust the isovector tensor coupling constant . Such a procedure exemplifies the focus of fit on the s.p. properties, as discussed in the Introduction.
It turns out that current experimental data, and in particular lack of information in Ni, do not allow for adjusting the fourth coupling constant, . For this reason, in the present study we fix it by keeping the ratio of equal to that of the given standard Skyrme force. In the process of fitting, all the remaining timeeven coupling constants are kept unchanged. Variants of the standard functionals obtained in this way are below denoted by SkP, SLy4, and SkO. When the timeodd channels, modified so as to reproduce the Landau parameters, are active, we also use notation SkP, SLy4, and SkO.
For the SkP functional, the procedure is illustrated in Fig. 3. We start with the isoscalar nucleus Ca. The evolution of the SO splittings in function of , which is the factor scaling the original SkP coupling constant , is shown in the upper panel of Fig. 3. As it is clearly seen from the Figure, fair agreement with data requires about 20% reduction in the conventional SO interaction strength , cf. results of the recent study in Ref. [38]. It should also be noted that the reduction in the SO interaction considerably improves the and splittings but slightly spoils the SO splitting. Qualitatively, similar results were obtained for the SLy4 and SkO interactions. Reasonable agreement to the data requires 20% reduction of the original in case of the SkO interaction and quite drastic 35% reduction of the original in case of the SLy4 force.
Having fixed in Ca we move to the isoscalar nucleus Ni. This nucleus is spinunsaturated and therefore is very sensitive to the isoscalar tensor coupling constant. The evolution of theoretical s.p. levels versus is illustrated in the middle panel of Fig. 3. As shown in the Figure, reasonable agreement between the empirical and theoretical SO splitting is achieved for MeVfm, which by a factor of about five exceeds the original SkP value for this coupling constant. It is striking that a similar value of is obtained in the analogical analysis performed for the SLy4 interaction.
Finally, the isovector tensor coupling constant is established in nucleus Ca. The evolution of theoretical neutron s.p. levels versus is illustrated in the lowest panel of Fig. 3. As shown in the Figure, the value of MeVfm is needed to reach reasonable agreement for the SO splitting in this case. For this value of the strength one obtains also good agreement for the proton SO splitting (see Figs. 4 and 5 below), without any further readjustment of the strength. Again, very similar value for the strength is deduced for the SLy4 force. Note also the improvement in the splitting caused by the isovector tensor interaction. Dotted lines show results obtained from the mass differences, i.e., with all the polarization effects included.
During the fitting procedure all the remaining functional coupling constants were kept fixed at their Skyrme values. The ratio of the isoscalar to the isovector coupling constant in the SO interaction channel was locked to its standard Skyrme value of . Since no clear indication for relaxing this condition is seen (see also Figs. 4 and 5 below), we have decided to investigate the isovector degree of freedom in the SO interaction (see [17]) by performing our threestep fitting process also for the generalized Skyrme interaction SkO [20], for which .
All the adopted functional coupling constants resulting from our calculations are collected in Table 4. Note, that the procedure leads to essentially identical SO interaction strengths for all three forces irrespective of their intrinsic differences, for example in effective masses. The tensor coupling constants in both the SkP and the SLy4 functionals are also very similar. In the SkO case, one observes rather clear enhancement in the isovector tensor coupling constant which becomes more negative to, most likely, counterbalance the nonstandard positive strength in the isovector SO channel.
The functionals were modified using only three specific pieces of data on the neutron SO splittings. In order to verify the reliability of the modifications, we have performed systematic calculations of the experimentally accessible SO splittings. The results are depicted in Figs. 4, 5, and 6 for the SkP, SLy4, and SkO functionals, respectively. Additionally, Fig. 7 shows neutron and proton magic gaps (24) calculated using the SkP functional. In all these Figures, estimates taken from Ref. [8] are used as reference empirical data.
This global set of the results can be summarized as follows:
Without any doubt the SO splittings are better described by the modified functionals. It should be stressed that the improvements were reached using only three additional data points without any further optimization. The tensor coupling constants deduced in this work and collected in table 4 should be therefore considered as reference values. Indeed, direct calculations show that variations in within 10% affect the calculated SO splittings only very weakly. The price paid for the improvements concerns mostly the binding energies, which for the nuclei Ni, Sn, and Pb become worse as compared to the original values. This issue is addressed in the next section.