Critical evaluation of Breit and QED effects on the $3d^{9}\ {}^{2}D_{3/2}\rightarrow^{2}D_{5/2}$ transition energy in Co-like ions.

The MCDHF method implemented in the GRASP2K package starts from a Dirac-Coulomb Hamiltonian $H_{DC}$

where $V_{i}=-\frac{Z}{r_{i}}$ is the monopole part of the electron-nucleus interaction, $r_{ij}$ the distance between electrons $i$ and $j$, and $\alpha$ and $\beta$ are the Dirac matrices. Electron correlation effect is included by expanding our Atomic State Function (ASF) $\Psi\left(\Gamma PJ\right)$ in a linear combination of Configuration State Functions (CSFs), $\Phi\left(\gamma_{i}PJ\right)$

where $\gamma_{i}$ represents all other quantum numbers needed to uniquely define the CSF. The CSFs are spin-angular coupled, antisymmetric products of Dirac orbitals of the form

The radial part of the one-electron orbitals and the expansion coefficients $c_{i}$ of the CSFs are obtained in the relativistic self-consistent field (RSCF) procedure. These are followed by a configuration interaction (RCI) approach, where Breit and QED effects are included. This implies that a limitation in the GRASP2K package could be that the effects that are only included in the RCI step do not affect the orbitals, since they are not included in the RSCF procedure. To investigate the importance of this, we went back to the results of a single-configuation approach (DHF), only including the $3d^{9}\ {}^{2}D_{3/2}$ and $3d^{9}\ {}^{2}D_{5/2}$ using the GRASP2K code and compared it to results from a B-spline version of a DHF program, DBSR-HF Zatsarinny2016, where the differential equations can be replaced by a set of generalized eigenvalue problems. In the latter code there are two options to include the Breit frequency-independent interaction, in the first one, it is added in the final stage as the GRASP2K code does, while in the second one, it is included into orbital optimization.

To represent correlation we start by using an extended version of the method from Guo et al. Guo2016 to include valence and core-valence correlation (where the $3d$ is defined as the only valence subshell). In this we allow for single and double excitations from the $3d$ subshell, as well as single excitation from all core subshells ($3p,3s,2p,2s,1s$), to an active set of orbitals with $l\leq 5$ and $n\leq 8$, to reach a clear convergence of electron correlation effect. The convergence trend of the energy splittings for the ions of interest here is given in Fig. 1. We can see that the energy splittings for the first 3 ions and the last 4 ions are converged to 0.015% and 0.0025%, respectively. We can conclude, since the correlation is relatively small in the Co-like ions, that the ”truncation” uncertainty in the computed fine structure, due to left out correlation, is negligible.

The Transverse Photon interaction is included in GRASP2K through a standard Hamiltonian, as a correction to order $\alpha^{2}$ to the Dirac-Coulomb Hamiltionian included in the RSCF calculation

where $\omega_{ij}$ is the frequency of the exchanged virtual photon. This reduces to the frequency-independent Breit interaction when $\omega_{ij}\rightarrow 0$ which we will label Breit(0). The remaining and frequency-dependent part, we will label Breit($\omega$). Breit(0) is the dominating correction to the Dirac-Coulomb results for all ions in the sequence and it is basically well understood how to include it in our calculations. This is not the case for the Breit($\omega$) part, since the frequency is only representing a physical property for spectroscopic orbitals, i.e. the orbitals $1s,2s,2p,3s,3p$ and $3d$ that are occupied in the reference term of $3d^{9}\ {}^{2}D$. We therefore choose a method where the frequency-dependent contribution is only evaluated between spectroscopic orbitals and put to zero if any others, often labeled correlation orbitals, are involved.

There are two contributions to the QED-correction, the Self-energy (SE) and the Vacuum Polarization (VP). For the Co-like ions, the SE dominates for all ions and we will therefore focus our investigation on different approaches to represent it. For the VP-contribtuion we evaluate the Uehling model potentials together with some higher order corrections Fullerton1976, as represented by the standard GRASP2K-approach.

In the current GRASP2K package, SE corrections are obtained based on a screened hydrogenic approximation

where $F(nlj,Z\alpha)$ is a slowly varying function of $Z\alpha$. The total self-energy correction is given as a sum of one electron corrections weighted by the fractional occupation number of the one-electron orbital in the wave function. In the current GRASP2K package, the SE correction was included by relying on the hydrogenic results of Mohr et al. Mohr1983 and Klarsfeld et al. Klarsfeld1973, the screening effect was included through a screened nuclear charge by taking the overlap integral of the wavefunction and a hydrogenic wavefunction. We will label the original GRASP2K calculation as ‘GRASP2K’.

By updating the GRASP2K program to use the latest available hydrogenic values Mohr1992; LeBigot2001 and modifying it to account for finite-nuclear-size effects Mohr1993; Beier1998, Lowe et al. Lowe2013 implemented a self-energy screening approximation based on the Welton interpretation Welton1948, it is labeled as ‘Welton’ in the following.

Recently, Shabaev et al. Shabaev2013; Shabaev2015 developed a model QED operator which also includes the non-local QED part to calculate the QED corrections for many-electron atomic systems. Two nuclear model types are supported, the point nucleus and the extended nucleus, where we always use the latter. We have included this model operator in the GRASP2K package and the results are labeled ’Shabaev’.

Since a possible limitation of the GRASP2K package is the fact that Breit is not included in the variational RSCF calculation, and therefore do not affect the computed orbitals, but only affect the wave function through the mixing between different states, we compare in Table 1 the DHF results from the GRASP2K with our DBSR-HF calculations. In the latter the frequency-independent Breit interaction is either computed as in GRASP2K in a non-variation fashion - results labelled B(1) - or included in the variational derivation of the orbitals in the method labelled B(2). It is apparent from Table 1 that the DHF energy splittings calculated using the GRASP2K code and DBSR-HF B(1) code are in excellent agreement, as expected. The ”variational effect” of the frequency-independent Breit, as manifested by the differences between the B(1) and B(2) options of DBSR ranges from 1 cm^-1 for Zr¹³⁺, to 67 cm^-1 for Au⁵²⁺. This is a rough or order of magnitude estimate of the uncertainty in our DHF calculations, introduced by not including the Breit operator in the RSCF-part of our calculations. It is reasonable to assume that the effect of this will decrease when correlation is included, since some of the effect will be incorporated in the expansion over CSFs. As we will see this variational effect is even for DHF an order of magnitude smaller than contributions from correlation and will therefore not affect the conclusions of this paper.

As discussed above, we use three different approximations to deal with the SE corrections, based on standard GRASP2K Jonsson2013, the Welton interpretation Welton1948; Lowe2013 and the Shabaev approach Shabaev2013; Shabaev2015. These are compared in Fig. 2, and it is clear that the results based on the Shabaev-method are larger than the GRASP2K results for all ions, and their differences smoothly vary with increasing atomic number. The SE contributions calculated using Welton’s concept are also larger than the GRASP2K results for $Z\leq 60$ but then decreases to become smaller for larger $Z$. There is a significant jump for the differences between these two calculations at $Z=60$, which is due to the fact that the SE contributions from electrons with $n=3,4,5$ and $5/2\leq j\leq 9/2$ for $60\leq Z\leq 110$ were taken from the results of Le Bigot et al. LeBigot2001, but those for $Z<60$ were obtained from an extrapolation.

Table 2 and Figure 3 present different contributions to the energy splitting for a number of ions, where direct measurements of the fine structure is available. It is clear that Breit(0) represents the largest contribution, followed by Self-energy, for most ions. Since the correlation and Breit($\omega$) have opposite signs for large Z, and almost cancel each other, these ions are excellent testing ground for the remaining Self-energy contribution.

In Table 3 and Figure 4 we present the final results for our three different models for Self-energy, compared with experimental observations. For the first three ions, the final results for all models are all within or close to the experimental error bars. However, for the four high-$Z$ ions, none of the three calculations agree with experiment to within the experimental error bar, albeit the Shabaev-method is closest, being lower than the experimental values by 0.03%$-$0.04%. It is clear that this can not be attributed to correlation, since for these four ions, the convergence of our present calculation is about 25 ppm and the left-out core-core (CC) correlation has en estimated, maximal contribution of only 30 ppm Guo2016a.

The final results for the whole isoelectronic squence with $28\leq Z\leq 100$ are listed in Table 4. Although except for the 7 ions where direct measurements of the fine structure is available, the active set for $28\leq Z\leq 42$ is restricted to $n\leq 7,l\leq 5$, while for $43\leq Z\leq 100$ is $n\leq 6,l\leq 5$, as shown in figure 1 and our previous paper Guo2016, the convergence of the electron correlation calculations is within 0.01%.