Johann V.
Pototschnig
*ab,
Kenneth G.
Dyall
c,
Lucas
Visscher
b and
André Severo Pereira
Gomes
d
aInstitut für Experimentalphysik, Technische Universität Graz, Petersgasse 16, 8010 Graz, Austria. E-mail: johann.pototschnig@tugraz.at
bDepartment of Chemistry and Pharmaceutical Sciences, Faculty of Sciences, Vrije Universiteit Amsterdam, De Boelelaan 1083, NL-1081 HV Amsterdam, The Netherlands. E-mail: visscher@chem.vu.nl
cDirac Solutions, 10527 NW Lost Park Drive, Portland, OR 97229, USA. E-mail: diracsolutions@gmail.com
dUniversité de Lille, CNRS, UMR 8523 - PhLAM - Physique des Lasers, Atomes et Molécules, F-59000 Lille, France. E-mail: andre.gomes@univ-lille.fr
First published on 1st October 2021
We report an investigation of the low-lying excited states of the YbF molecule-a candidate molecule for experimental measurements of the electron electric dipole moment-with 2-component based multi-reference configuration interaction (MRCI), equation of motion coupled cluster (EOM-CCSD) and the extrapolated intermediate Hamiltonian Fock-space coupled cluster (XIHFS-CCSD). Specifically, we address the question of the nature of these low-lying states in terms of configurations containing filled or partially-filled Yb 4f shells. We show that while it does not appear possible to carry out calculations with both kinds of configurations contained in the same active space, reliable information can be extracted from different sectors of Fock space-that is, by performing electron attachment and detachment IHFS-CCSD and EOM-CCSD calculation on the closed-shell YbF+ and YbF− species, respectively. From these calculations we predict Ω = 1/2, 3/2 states, arising from the 4f13σ26s, 4f145d1/6p1, and 4f135d1σ16s configurations to be able to interact as they appear in the same energy range around the ground-state equilibrium geometry. As these states are generated from different sectors of Fock space, they are almost orthogonal and provide complementary descriptions of parts of the excited state manifold. To obtain a comprehensive picture, we introduce a simple adiabatization model to extract energies of interacting Ω = 1/2, 3/2 states that can be compared to experimental observations.
A noteworthy finding in our previous work was the sensitivity of coupled cluster calculations to the basis set in use and, indirectly, to the amount of electron correlation recovered in the calculations. We observed a spike in the values of the T1 diagnostic around the ground state equilibrium geometry, so it appears that the perturbative treatment of triple excitations in the CCSD(T) calculations breaks down in this region of the potential energy curves unless there is enough flexibility in the correlation treatment. The same was recently observed by Pasteka et al.16 for the nuclear quadruple coupling constant. This suggests the existence of a low lying perturbing state, which we want to investigate further in the current work.
Experimental17,18 and previous theoretical19–21 investigations suggest that in the ground state the unpaired electron is located in a σs orbital with dominant contributions from the 6s orbital of Yb, corresponding to a Yb(4f14σ16s)F configuration. This 2Σ1/2 state ground state was studied in greater detail by combining microwave and optical spectroscopy for the odd 171Yb isotope.22
Experimentally,18 the lowest excited state observed is assigned as 2Π1/2, with an energy of 18106 cm−1, while the 2Π3/2 component is found at 19
471 cm−1, yielding a spin–orbit splitting of 1365 cm−1 of this spin–orbit split A 2Π state. The lower component will be denoted 31/2 in the current work. Experiments indicate a perturbation of its vibrational levels,18,23–25 which was attributed to the presence of a perturbing state (denoted by 41/2 here) found at 18
705 cm−1.18 This perturbing 41/2 state is sometimes referred to as [18.6]1/2 by experimentalists18,23 (energy in cm−1 divided by 103 in the square brackets, and Ω-value as subscript). The mixing of these two Ω = 1/2 states gives rise to states designated as [557] and [561] (the values in square brackets referring to transition energies in Thz from the vibronic ground state) with transition energies of 18
574 and 18
699 cm−1, respectively.24,25 These two states are of importance for laser cooling schemes that have been investigated26 and tested25 with the purpose of realizing high-accuracy measurements of YbF at very low temperatures. Besides these first excited states, Smallman26 investigated also two not yet fully characterized mixed states, [574](≈19
150 cm−1) and [578](≈19
280 cm−1) at higher energies. These can be compared with the Ω = 3/2 state at 19
471 cm−1 found earlier by Dunfield,18 which will be denoted 23/2 in the current work. Uttam et al.27,28 furthermore measured additional unidentified higher bands at about 23
035, 23
256 and 26
015 cm−1, which they denoted as C1, C2 and D, respectively.
Theoretically, excited states arising from the Yb(4f146p1)F and Yb(4f145d1)F configurations were considered by Nayak and Chaudhuri6 with RAS-CI based on 4-component spinors, yielding the A2Π1/2 (31/2), A2Π3/2 (23/2), and a 2Σ1/2 state. Earlier multireference CI calculations by Dolg et al.19 furthermore indicate the possibility of low-lying Ω = 1/2 states arising from the or
configurations, to be lying below or close to the Yb(4f146p1)F states. This was also found in the DFT calculations of Liu et al.20 who place excited states arising from the
configuration in the range from 9000 to 15
000 cm−1 relative to the Yb(4f14σ16s)F ground state. These findings make it of interest to explicitly consider the configuration interaction between the f13 and f14 configurations in the Yb atom.29
The vibronic states are additionally split due to hyperfine interactions. In atomic experiments they were measured for the ground and excited states9,22,30–33 also using Zeeman spectroscopy.34 The hyperfine interaction of the atom35–37 and molecule38 were studied theoretically, and should have similar uncertainties to the contribution of the eEDM to the spectrum due to the similarity of the matrix elements. Recently, uncertainties of the hyperfine constants arising in relativistic coupled cluster computations have been studied.39
It is clear from the above that a proper description of the Yb atom and the YbF molecule requires an accurate treatment of both spin–orbit coupling and electron correlation, for ground as well as excited states. A popular approach is the so-called two-step approach to spin–orbit coupling (SOC), in which electron correlation methods based on non-relativistic or scalar relativistic Hamiltonians are used to obtain excited state energies, that are in turn used to dress a spin–orbit configuration interaction (SOCI) matrix. This approach can yield quite accurate spin–orbit coupled states, but results are particularly sensitive to the number of spin-free states serving as a basis for the SOCI step.40–46 An alternative is to include SOC already at the mean-field level, and use fully SO-coupled molecular spinors to construct the correlated wave functions.47 This can be done with four-component Hamiltonians, as done for the ground1 and excited states6,37,48,49 of YbF, or with more computationally efficient two-component Hamiltonians based on the eXact 2-Component (X2C) approach,50–57 in which a transformation to decouple the positive and negative energy states of the Dirac Hamiltonian can be carried out in matrix form, yielding the same positive energy spectrum as the original 4-component Hamiltonian. More details can be found in the recent review by Liu.47 Among the different X2C flavors, we can distinguish two main strategies for the decoupling, which is performed based on: (i) the one-electron Dirac Hamiltonian prior to the mean-field step,54,57 and for which two-electron spin–orbit contributions due to the untransformed two-electron potential are included via atomic mean-field contributions calculated with the AMFI code41,58,59 (X2C-AMFI); (ii) after a converged 4-component mean-field calculation on atoms52,53,56 or molecules55 (2DCM). Recent benchmarks show that 2DCM calculations closely reproduce equivalent 4-component ones for valence60 or core61 states.
Moreover, the aforementioned calculations for the excited states of YbF have mostly employed multireference CI (MRCI) approaches. While these can provide great flexibility in capturing static correlation, it remains the case that dynamical correlation is better accounted for with coupled cluster approaches. Among the coupled cluster singles and doubles (CCSD) approaches for excited states, we have the equation of motion (EOM-CCSD) method as well as Fock-space (IHFS-CCSD) methods,62 of which the single electron attachment, detachment, and singly excited states variants are the most commonly used. The two approaches have been found to yield very accurate results in general and in particular for calculations with relativistic Hamiltonians as discussed elsewhere (see ref. 60 and references therein).
The first goal of this work is therefore to go beyond the investigations performed to date in the literature, and apply the relativistic EOM-CCSD and IHFS-CCSD approaches to describe the low-lying excited states of YbF. For such states, where the most important excited state configurations appear have a single open-shell character (4f14σ16s, 4f145d1,
4f146p1), these coupled cluster approaches are in principle applicable, provided one starts from closed shell configurations such as Yb(4f14σ26s)F− or Yb(4f14)F+. Additionally, we assess the performance of relativistic MRCI with respect to the coupled cluster methods. Our second goal is to confirm whether any low-lying state is close enough to the ground state to perturb the latter, and explain the anomalous behavior observed in the open-shell ground-state calculations in the literature.
![]() | (1) |
In the coupled cluster computations the 2DCM Hamiltonian54,55 was applied, all two-electron integrals over small component (S) basis sets (i.e. the so-called (SS|SS)-type integrals) appearing in the SCF step have been replaced by a simple correction.66 In order to account for spin–orbit coupling and other relativistic effects the X2C-AMFI Hamiltonian was employed for the Kramers-restricted configuration interaction (KRCI) method.
Spectroscopic constants (re, De, ωe and ωeχe) were determined from a Morse potential fit in the vicinity of the potential energy minima. The potential energy curves were determined for bond lengths between 1.6 Å and 2.3 Å spaced by 0.02 Å and additional points with larger spacing up to 3.5 Å. In the calculation of De the asymptotic dissociation limit is calculated from the energies of the isolated neutral atoms, F in the 2P3/2 state and Yb in the 1S0 state.
The dataset associated with this manuscript (outputs from calculations, codes to extract and process information from these, and code to obtain the spectroscopic constants) is provided in ref. 67.
The first set of states was obtained by electron attachment on Yb(4f14F+ ion, where the HOMO (σ6s,1/2) of YbF was initially empty. This computation on the (0h,1p) sector of Fock space yielded states with 4f14 and a valence electron in the σ6s, d or p orbital. This means that, in the process of obtaining the potentials for the ground and excited states of YbF, we immediately obtain energies of CCSD quality for YbF+, and therefore vertical ionization potentials (IP) at each geometry.
Another set of states was obtained by ionizing the Yb(4f14σ26s)F− anion, where the HOMO (σ6s,1/2) of YbF was initially doubly occupied. States arising from the Yb(4f14σ16s)F, and Yb(4f14σ26s)F(2p5) configurations were obtained by considering the (1h,0p) sector of Fock space. This means that, in the process of obtaining the potentials for the ground and excited states of YbF, we immediately obtain energies of CCSD quality for YbF−, and therefore vertical electron affinities (EA) at each geometry. We note that states arising from the (2h,1p) and (1h,2p) manifolds are also accessible from EOM-IP and EOM-EA calculations, though the energy of electronic states determined by such configurations will be less accurate than states dominated by single detachment or attachment configurations.
The EOM-CCSD electronic states are obtained by an iterative diagonalization (Davidson) procedure in which only the energies of a certain number of the lowest states are determined. For the IP-EOM-CCSD we obtained 16 Ω = 1/2, 8 Ω = 3/2, 6 Ω = 5/2 and 2 Ω = 7/2 states, whereas for EA-EOM-CCSD we obtained 8 Ω = 1/2, 6 Ω = 3/2, 4 Ω = 5/2 and 2 Ω = 7/2 states.
As transition moments are not yet available for the EOM-CCSD implementation in Dirac, we have only obtained the potential energy curves. These are nevertheless useful since, by not requiring the definition of model spaces or the use of an extrapolation procedure, they serve as a cross validation of the IHFS-CCSD calculations below.
The separation into a model and external space leads to the appearance of the so-called intruder states, a well-known difficulty with Fock-space coupled cluster and other effective Hamiltonian approaches, that can be dealt with in many cases by the intermediate Hamiltonian (IH) Fock-space coupled cluster (IHFS-CCSD) method.71,72
The IH approach was employed to compute Yb(4f14{σ6s,6p,5d…}1)F states starting from YbF+. The active P space in such calculations contained about 50 spinors varying slightly with bond distance and basis set. Of these 26 spinors are always present in the model (Pm) space, whereas the remaining active spinors are placed in the intermediate (PI) space. Due to using the (0h,1p) sector for the cation, states arising from configurations where the Yb 4f shell is partially filled (such as or
) are not accessible in this calculation.
The approach outlined above was, however, not enough to avoid divergence in the computation of the (1h,0p) sector using the anion as a reference. Therefore, the extrapolated intermediate Hamiltonian (XIH) Fock-space coupled-cluster approach73 (XIHFS-CCSD) was applied using the same shifts as in ref. 73. Values of 0.1 and 0.2 Hartree were selected if one of the holes is not in the model space. These shifts were doubled for two holes outside the model space. Using the determined energies an extrapolation to the system without shifts was performed. The model (Pm) space in these computations contained 22 spinors, the intermediate (PI) space about 24 spinors depending on the bond distance and basis set. Since we start out from the anion and only allow holes, only Yb(4f14σ16s)F, and Yb(4f14σ26s)F(2p5) configurations are accessible in this computation.
Combining the two sectors allows us to get different excited states of YbF, although there are limitations. Firstly, the interactions between configurations with open f-shell and the ones with an electron in the p- or d-shell are not included, since they will be obtained for different sectors of Fock space. This interaction will nevertheless be treated with a simple adiabatization approach, described in Sections 2.4 and 3.6. Secondly, configurations such as or
are not included in the current treatment. This limitations is not as significant because these states have higher energies than the ones we are interested in. Both of these problems could be dealt with by using the (1h,1p) sector, but this goes beyond the current work as convergence is very unstable for this sector and it requires the use of an open-shell reference.
![]() | (2) |
Subsequently, the potential energy curves for Kramers-restricted configuration interaction are presented. A discussion of the coupled cluster approaches follows, with a focus on the comparison of the coupled cluster results for the Fock space and equation-of-motion approach. This section is followed by a presentation of the spectroscopic data for the ground and excited states. In the last part we take a closer look at the mixing of states at around 18000 cm−1 and apply the adiabatization procedure.
These observations help to understand why it turned out to be very difficult to treat both sets of states in the same calculation, which we attempted to do from AOC-SCF on the Yb+. We started out by performing AOC-SCF computations on the atom, based on the 4f closed shell configuration. While we obtained the correct ground state configuration, the has an energy of about 46
000 cm−1 (over two times higher than the experimental value), and the wrong order for the hole states is observed. If the wave function is optimized for a 4f13 configuration, one obtains the
as the lowest state and finds the true ground state more than 20
000 cm−1 higher. Because of these difficulties, the KRCI calculations discussed below all follow the strategy of different orbital sets that is also employed in the subsequent coupled cluster calculations.
Table 1 contains KRCI values of electronic transitions for the cation. The transition energies show deviations of about 10% and the spin–orbit splitting is underestimated for states with a 4f14 configuration. The squared transition dipole moment (TDM2) of the state is underestimated by about 13%, the one for the
state overestimated by about 47%. The second set of states with a hole in the f-shell and different distributions of the 2 valence electrons are given in the lower part of Table 1, the energies are relative to the
state. In this case the two valence electrons are distributed over the s- and d-shell. The lowest state with a
configuration has a transition energy of 47912.31 cm−1 and was not included in the current treatment.
State | Conf. | NIST74 | 2z | 3z | 4z | extr. | ||||
---|---|---|---|---|---|---|---|---|---|---|
E | S | E | TDM2 | E | TDM2 | E | TDM2 | E | ||
2S1/2 | 4f146s1 | 0 | 0 | 0 | 0 | 0 | ||||
2D3/2 | 4f145d1 | 22![]() |
23![]() |
0.0 | 22![]() |
0.0 | 23![]() |
0.0 | 24![]() |
|
2D5/2 | 4f145d1 | 24333 | 23![]() |
0.0 | 23![]() |
0.0 | 24![]() |
0.0 | 24![]() |
|
![]() |
4f146p1 | 27062 | 6.1 | 25![]() |
3.5 | 24![]() |
3.8 | 25![]() |
3.6 | 25![]() |
![]() |
4f146p1 | 30392 | 11.4 | 28![]() |
16.9 | 27![]() |
18.9 | 28![]() |
17.4 | 28![]() |
The excited states of the Yb atom for the 4f14 configuration have already been investigated by relativistic Fock-space coupled cluster36,37,75,76 as well as for the cation,75 including the transition moment of magnetic transitions.76 With our current calculations we can go beyond these studies and investigate the f13 configurations as well. Before discussing our IHFS-CCSD calculations for Yb+, we focus on the EOM-CCSD excitation energies, shown in Table 2. The EOM-IP-CCSD energies of 4f13 states obtained from the extrapolation to the complete basis set limit underestimate the experimental transition energies by around 3000 cm−1, whereas the values for 4f14 states, obtained with EOM-EA-CCSD are within 1000 cm−1 of the experimental values, which yields a quantitative improvement over the KRCI ones for both configurations, even though qualitatively the two methods provide a similar picture. From that and the preceeding discussion, we attribute the relatively lower accuracy for the 4f13 to arise from the incomplete account of the relaxation of the wave function upon the creation of the hole in the f shell. Beyond the states presented in Table 2, which are dominated by single electron attachment and detachment, we are able to access states with significant (1h,2p) and (2h,1p) character with EOM-CCSD. These states, available in Table S14 in the ESI,† are about 10000 cm−1 higher in energy than the experimental ones.
State | Conf. | NIST74 | 2z | 3z | 4z | Extr. |
---|---|---|---|---|---|---|
2S1/2 | 4f146s1 | 0 | 0 | 0 | 0 | 0 |
![]() |
4f136s2 | 21![]() |
12![]() |
13![]() |
16![]() |
17![]() |
![]() |
4f136s2 | 31![]() |
22![]() |
24![]() |
26![]() |
28![]() |
2D3/2 | 4f145d1 | 22![]() |
24![]() |
24![]() |
24![]() |
23![]() |
2D5/2 | 4f145d1 | 24![]() |
25![]() |
25![]() |
25![]() |
25![]() |
![]() |
4f146p1 | 27![]() |
27![]() |
27![]() |
27![]() |
27![]() |
![]() |
4f146p1 | 30![]() |
30![]() |
31![]() |
31![]() |
31![]() |
Finally, our IHFS-CCSD results are presented in Table 3. The transition energies for 4f14 configuration reproduce well the experimental ones, with errors below 6%, and only show a small dependence on the basis set. The states arising from the 4f13 configuration (etc.), in contrast, show a significant dependence on the basis and a rather slow convergence and underestimate the value by about 30%, which makes them less accurate than the EOM-CCSD ones. This lower accuracy is a consequence of the reduced flexibility in the model spaces, due to the need of adding the 5p-shell just below the 4f-shell to the intermediate space, in order to achieve convergence. These results are in line with the observations of Shee et al.,60 in that the formal equivalence between EOM-CCSD and IHFS-CCSD for the sectors of Fock space considered depends, in fact, on the flexibility of the main model space.
State | Conf. | NIST74 | 2z | 3z | 4z | Extr. | DCB75 |
---|---|---|---|---|---|---|---|
2S1/2 | 4f146s1 | 0 | 0 | 0 | 0 | 0 | 0 |
![]() |
4f136s2 | 21![]() |
11![]() |
12![]() |
13![]() |
14![]() |
|
![]() |
4f136s2 | 31![]() |
21![]() |
22![]() |
24![]() |
25![]() |
|
2D3/2 | 4f145d1 | 22![]() |
24![]() |
24![]() |
24![]() |
23![]() |
23![]() |
2D5/2 | 4f145d1 | 24![]() |
25![]() |
25![]() |
25![]() |
25![]() |
24![]() |
![]() |
4f146p1 | 27![]() |
27![]() |
27![]() |
27![]() |
27![]() |
27![]() |
![]() |
4f146p1 | 30![]() |
30![]() |
31![]() |
31![]() |
31![]() |
31![]() |
Furthermore, the removal of the 5p spinors from the main model space underscores the importance of the 5p for the energetics of the states with a hole in the 4f shell, since by doing so, we undress the contributions from the 5p configurations, and thus prevent them from interacting effectively with 4f13 determinants.
![]() | ||
Fig. 1 Combination of the sets of KRCI potentials obtained by extrapolating triple and quadruple zeta basis sets. The lowest Ω = 1/2 states are denoted by their dominant configuration. |
The potential energy curves have been determined up to 15 Å for the closed and open f-shells. The energy difference in the atom between these states is 21418.75 cm−1. Accordingly, the PECs for the hole states were shifted to obtain this separation at this distance. There is still some interaction between ytterbium and fluorine at 15 Å, but the long range behaviour can be expected to be similar for the two configurations (this assumption was checked, see Fig. S8 in the ESI,† for further details). Taking into account the position of the minima, the curvatures, spin–orbit splitting, the avoided crossings and asymptotes the states can be assigned to a dominant configuration, shown in Fig. 1.
Regarding the 4f14 manifold, the lowest two excited states in the figure belong to the Yb(4f146p1)F configuration, but approach asymptotically the 2D3/2 state. The asymptote of the next three states is 2D5/2 corresponding to the Yb(4f145d1)F configuration for smaller internuclear separations. For Ω = 5/2 the transition dipole moment with the ground state is zero, for the other four the values are shown in Fig. S3 in the ESI.† The first Ω = 3/2 and the third Ω = 1/2 state have a larger transition dipole moment close to equilibrium, but get close to each other at the largest internuclear separations.
Regarding states of the 4f13 manifold, the lowest four states belonging to the configuration are well separated from a dense region with a lot of states about 12
000 cm−1 higher. Most of these states are of the
configuration, with the
state slightly higher in energy asymptotically and more strongly bound, resulting in several avoided crossings. For each of the four
states the transition dipole moments with higher excited states of varying Ω are plotted in Fig. S6 in the ESI.† The transition dipole moments are substantially smaller than the ones for the closed 4f-shells but some of them are non-zero.
An alternative to AOC-SCF for obtaining orbitals for several configurations is multiconfigurational SCF, but similar difficulties as for AOC-SCF in obtaining a balanced description of the 4f14 and 4f13 states are observed: either the wrong ground state is obtained (if only the hole states are optimized in MCSCF), or the hole states are too high in energy by about 20000 cm−1 (if the ground state is optimized). We also made attempts using state-averaged MCSCF in a non-relativistic quantum chemistry code and observed the same difficulties (see dataset67). If the 4f13 configurations are excluded one obtains meaningful results, but at the expense of obtaining a
states too high in energy. If all the states are included, the wrong ground state is obtained.
![]() | ||
Fig. 2 Potential energy curves obtained by extrapolating triple and quadruple zeta basis sets. EOM-CCSD results in the upper part, IHFS-CCSD in the lower part. The Ω values of 1/2, 3/2, 5/2 and 7/2 are indicated by the colors red, blue, green and orange. The light colors are used for the (0h,1p) sector, dark ones for (1h,0p). For the states below 30![]() |
While the EOM-CCSD excitations energies of Yb+ are closer to the experimental ones, the states are too high to perturb the Yb(4f146p1/5d1)F PECs. From the extended potential energy curves provided in the ESI† (Fig. S15 and S16), we can observe that the ground state of the non-interacting system (Yb(4f14σ26s)F(2p5)) is repulsive and has a high energy at the equilibrium distance. This results in several avoided crossings being observed at 3, 3.5, and 5 Å.
Since the Yb(4f14σ16s)F ground state is accessible for both sectors employed in the coupled cluster calculations (Fock space as well as EOM), we can assess the compatibility of the two separate sets of calculations (in the sense of having comparable accuracies) by looking more closely at the differences between the ground states in Fig. 2. From that, we can see that the EOM-CCSD and IHFS-CCSD approaches the curves are on top of each other from the smallest considered internuclear separation up to about 2.8 Å. This assures us that there should not be artifacts in putting together and comparing the calculations on the two sectors.
Quant. | System | KRCI | EOM-CCSD | IHFS-CCSD | Experiment | |||
---|---|---|---|---|---|---|---|---|
4z | Extr. | 4z | Extr. | 4z | Extr. | |||
IP | Yb | 38![]() |
39![]() |
50![]() |
50![]() |
50![]() |
50![]() |
50![]() |
IP2 | Yb | 90![]() |
90![]() |
97![]() |
98![]() |
97![]() |
98![]() |
98![]() |
IP | F | 127![]() |
126![]() |
144![]() |
143![]() |
144![]() |
144![]() |
140![]() |
EA | F | 13![]() |
11![]() |
27![]() |
27![]() |
27![]() |
27![]() |
27![]() |
IP | YbF | 58![]() |
56![]() |
48![]() |
48![]() |
48![]() |
49![]() |
47![]() |
EA | YbF | 7423 | 7326 | 9713 | 9876 | 9579 | 8197 | |
D e(IP) | YbF | 26![]() |
25![]() |
43![]() |
47![]() |
40![]() |
40![]() |
43![]() |
D e(EA) | YbF | 40![]() |
39![]() |
45![]() |
49![]() |
40![]() |
49![]() |
43![]() |
Unlike coupled cluster calculations, for KRCI a consistent definition of active spaces is difficult, and its lack of size-consistency results in large deviations from experiment and from the coupled cluster values. For adiabatic electron affinities, for which to the best of our knowledge there are no experimental values, the extrapolated values are 8393 and 8197 cm−1 for EOM-CCSD and IHFS-CCSD, respectively. For a distance of 6.5 Å a value of 28651 cm−1 was obtained, which is reasonably close to the electron affinity of fluorine (27
432 cm−1).77 Corresponding results for the atoms are listed in the table, which allow to calculate the dissociation energies (De). They deviate from the experimental values of 43
600 ± 800 cm−1 by Kaledin et al.78 and 43
260 ± 800 cm−1 by Yokozeki and Menzinger.79 The ionization potentials in Table 4 show acceptable agreement with experimental values.
Method | Ref. | r e | ω e | ω e χ e | D e |
---|---|---|---|---|---|
KRCI | YbF | 2.0829 | 465 | 2.40 | 39 660 |
EOM-CCSD | YbF+ | 2.0230 | 511 | 2.80 | 49![]() |
YbF− | 2.0250 | 508 | 2.53 | 47![]() |
|
IHFS-CCSD | YbF+ | 2.0176 | 515 | 2.82 | 49![]() |
YbF− | 2.0159 | 513 | 2.42 | 40![]() |
|
CCSD1 | YbF | 2.0174 | 507.6 | 2.357 | 40![]() |
CCSD(T)1 | YbF | 2.0289 | 528.2 | 1.939 | 41![]() |
CCSD21 | 2.0127 | 566.8 | 3.7885 | 55![]() |
|
RASCI6 | 2.051 | 529 | |||
CCSD(T)80 | 2.03 | 38![]() |
|||
CISD81 | 2.034 | 502 | 42![]() |
||
DFT20 | 1.987 | 532 | 45![]() |
||
Exp.79 | 43![]() |
||||
Exp.18 | 2.0158 | 506.6674 | 2.2452 | ||
Exp.27 | 505.5 | 1.9 | |||
Exp.78 | 43![]() |
||||
Exp.30 | 2.016514 | ||||
Exp.24 | 2.0195 | 506.616 | 2.235 |
We observe that the extrapolated KRCI bond distances, at about 2.058 Å, are significantly longer (by around 0.04 Å) than experiment,30 whereas the coupled cluster calculations show differences from experiment smaller than 0.01 Å, with EOM-CCSD showing slightly larger discrepancies than IHFS-CCSD. Between the extrapolated EOM-CCSD and IHFS-CCSD, we also see small differences between the 4f14 and 4f13 for EOM-CCSD these differ by around 0.001 Å whereas for IHFS-CCSD the difference is slightly under 0.002 Å, with the 4f13 configuration yielding a slightly underestimated value, compared to experiment, something that can be traced back to the differences in model spaces for this configuration.
Our results for harmonic frequencies further indicate that KRCI seems to underestimate the bonding strength in YbF, as the harmonic vibrational frequency is smaller (491 cm−1) than experiment (between 505.5 and 506.7 cm−1 depending on the experiment). The coupled cluster results, on the other hand, show the typical 5–6 cm−1 overestimation of the harmonic frequencies with respect to experiment (something also seen for the anharmonic constants), which can be attributed to lack of triples in the EOM or FS treatment, that would introduce further orbital relaxation. This can be seen in comparison to the unrestricted coupled cluster calculations of Gomes et al.,1 which in spite of the large value of the T1 diagnostic, reproduce well the experimental bond lengths, harmonic frequencies and anharmonic constants.
Taken together, our 2-component CCSD-based calculations and the 4-component ones of Gomes et al.1 compare consistently better to experiment than the other theoretical works for bond lengths, vibrational frequencies and anharmonic constants. For the dissociation energies, on the other hand, the extrapolated calculations presented here do not provide a significant improvement over the results of prior theoretical investigations (quadruple zeta values are closer to the experimental ones for this quantity, see Table 4). Especially, electron attachment values are off, which might be related to the absence of the configuration with a hole in the p orbitals of fluorine.
Moving now to excited states, we start by considering the four lowest excited states, which belong to the configuration. These states are well separated from the ground state (the lowest excited state is about 10
000 cm−1 above the ground state) and higher excited states. That such states are quite well separated from the ground state would, in our view, tend to exclude the interaction with a low-lying excited state as an explanation for the appearance of the large T1 diagnostic values observed by Gomes et al.1 From their spectroscopic constants, presented in Table 6, we see that with the exception of DFT all methods yield similar level splittings of about 500, 1200, and 2000 cm−1. To the best of the authors knowledge there is no experimental data available for these states, due to their negligible transition dipole moments for dipole excitations (see for instance Fig. S6 in the ESI†) and small Franck–Condon factors due to the difference in bond distances between these states and the ground state, see Section 3.6.
State | Method | T e | r e | ω e | ω e χ e | T rel |
---|---|---|---|---|---|---|
21/2 | KRCI | 1.9200 | 631 | 2.51 | ||
EOM-CCSD | 12![]() |
1.9432 | 591 | 2.59 | ||
IHFS-CCSD | 9627 | 1.9396 | 599 | 2.79 | ||
DFT20 | 3790 | 1.9570 | 561 | |||
MRCI19 | 1.9480 | 600 | ||||
13/2 | KRCI | 1.9253 | 628 | 2.50 | 540 | |
EOM-CCSD | 13![]() |
1.9494 | 588 | 2.61 | 643 | |
IHFS-CCSD | 10![]() |
1.9438 | 595 | 2.79 | 553 | |
DFT20 | 9520 | 1.9440 | 597 | 5730 | ||
MRCI19 | 1.951 0 | 598 | 428 | |||
15/2 | KRCI | 1.9296 | 622 | 2.45 | 1223 | |
EOM-CCSD | 13![]() |
1.9553 | 582 | 2.61 | 1135 | |
IHFS-CCSD | 10![]() |
1.9493 | 589 | 2.78 | 1341 | |
DFT20 | 10![]() |
1.9360 | 598 | 7180 | ||
MRCI19 | 1.9540 | 594 | 1021 | |||
17/2 | KRCI | 1.9315 | 616 | 2.43 | 1933 | |
EOM-CCSD | 14![]() |
1.9556 | 577 | 2.62 | 2117 | |
IHFS-CCSD | 11![]() |
1.9496 | 583 | 2.77 | 2018 | |
DFT20 | 16![]() |
1.936 | 592 | 12![]() |
||
MRCI19 | 1.954 | 589 | 1709 |
The smallest equilibrium distance was obtained for the 21/2 state with 1.94 Å for the coupled cluster methods and 0.02 Å less for KRCI. The vibrational frequencies are between 570 and 600 cm−1 for the coupled cluster methods and about 30 cm−1 higher for KRCI.
For higher excited states, as apparent from the figures in the previous section, the identification and assignment of states gets more difficult and there are differences between the methods. We have nevertheless provided in Table 7 the spectroscopic constants for excited with Ω values of 1/2, 3/2, and 5/2, respectively.
Ω | Method | State | Configuration | T e | r e | ω e | ω e χ e |
---|---|---|---|---|---|---|---|
a KRCI transition energies for the 4f13 sector were obtained by adding 4144 cm−1, an estimate for the energy of the lowest state in this manifold. | |||||||
1/2 | KRCIa | 3 | 4f13σ26s | 15![]() |
1.9038 | 655 | 13.57 |
5 | 4f146p1 | 16![]() |
2.0504 | 496 | 2.38 | ||
10 | 4f145d1 | 19![]() |
2.0552 | 490 | 2.49 | ||
EOM-CCSD | 3 | 4f146p1 | 18![]() |
2.0004 | 536 | 2.72 | |
4 | 4f145d1 | 21![]() |
2.0079 | 532 | 2.78 | ||
6 | 4f13σ26s | 23![]() |
1.9432 | 582 | 4.06 | ||
IHFS-CCSD | 3 | 4f146p1 | 18![]() |
1.9953 | 539 | 2.63 | |
4 | 4f13σ26s | 20![]() |
1.9397 | 597 | 2.78 | ||
5 | 4f145d1 | 21![]() |
2.0032 | 533 | 2.73 | ||
MRCI19 | 4f13σ26s | 1.948 | 600 | ||||
Exp.18![]() |
3 | 18106.20 | 537 | 3 | |||
Exp.18 | 4 | [18.6]1/2 | 18705.06 | ||||
Exp.24 | [557] | 18![]() |
1.9656 | 502.15 | |||
Exp.24 | [561] | 18![]() |
1.9571 | ||||
3/2 | KRCIa | 2 | 4f13σ26s | 16![]() |
1.9331 | 711 | 8.45 |
4 | 4f146p1 | 17![]() |
2.0473 | 499 | 2.37 | ||
>7 | 4f145d1 | 24![]() |
2.0669 | 470 | 2.50 | ||
EOM-CCSD | 2 | 4f146p1 | 19![]() |
1.9971 | 540 | 2.72 | |
4 | 4f13σ26s | 24![]() |
1.9537 | 584 | 2.64 | ||
5 | 4f145d1 | 24![]() |
2.0177 | 509 | 2.78 | ||
IHFS-CCSD | 2 | 4f146p1 | 19![]() |
1.9920 | 542 | 2.63 | |
3 | 4f13σ26s | 21![]() |
1.9480 | 591 | 2.80 | ||
4 | 4f145d1 | 24![]() |
2.0120 | 512 | 2.73 | ||
MRCI19 | 4f13σ26s | 1.953 | 596 | ||||
Exp.18 | 2 | 19471.49 | |||||
5/2 | KRCIa | 3 | 4f13σ26s | 17![]() |
1.9302 | 635 | 1.41 |
>6 | 4f145d1 | 24![]() |
2.0639 | 474 | 2.48 | ||
EOM-CCSD | 2 | 4f13σ26s | 24![]() |
1.9536 | 577 | 2.62 | |
3 | 4f145d1 | 25![]() |
2.0146 | 513 | 2.80 | ||
IHFS-CCSD | 2 | 4f13σ26s | 22![]() |
1.9499 | 584 | 2.77 | |
3 | 4f145d1 | 24![]() |
2.0089 | 515 | 2.77 | ||
MRCI19 | 4f13σ26s | 1.954 | 590 | ||||
n.a. | Exp.26 | [574] | 19![]() |
||||
Exp.26 | [578] | 19![]() |
|||||
Exp.27 | C1 | 23035.3 | 523 | 2 | |||
Exp.27 | C2 | 23256.0 | 507 | 2 | |||
Exp.27 | D | 26014.8 | 574.6 | 2.8 |
The comparison with experimental results allows assignment of the lowest excited state reported in experiments and give some indications for higher states. The lowest Ω = 1/2 state observed in experiment can be identified as the 31/2 state (51/2 for KRCI). Spectroscopic parameters agree well with the ones obtained by fitting to the A2Π1/2 in experiments. A bond distance of 1.9935 Å obtained by fitting to the same states in ref. 34 agrees well with the coupled cluster values for the 31/2 state, the vibrational constant of about 540 cm−1 is close to the experimental value of ref. 18. Similarly, the lowest Ω = 3/2 state reported by Dunfield et al.18 can be identified as the 23/2 state (43/2 for KRCI), see Table 7.
The lowest states with Ω = 1/2 and Ω = 3/2 for this energy range approach asymptotically a state with a Yb(4f145d1)F configuration, but if one analyses the EOM-CCSD and IHFS-CCSD orbital composition, significant contributions of the atomic 6p are identified. The Ω = 3/2 state is dominated (97%) by a single configuration, corresponding to a HOMO(σ6s,1/2) → LUMO+1 where the latter is made up of a mixture of 6pπ and 5dπ orbitals (the 6pπ contributions being the dominant − ≃80% – in the reference YbF+ orbitals). The few other significant configurations arise from excitations to higher-lying orbitals with increasingly large (≥50% 5dπ) contributions. The Π1/2 state is also dominated by a single configuration, now corresponding to a HOMO(σ1/2) → LUMO transition, and shows a rather similar picture in terms of the relative weights of the 6pπ and 5dπ orbitals, with very small contributions from the ground-state mixing due to spin–orbit coupling. The splitting in Yb+ of 2D3/2 and 2D5/2 is 1372 cm−1, for 2P1/2 and 2P3/2 it is 3330 cm−1. The separation between the lowest excited Ω = 1/2 states in the closed shell computation is 3779 and 3126 cm−1 for EOM-CCSD and IHFS-CCSD, respectively. This is an indication that the two states must be regarded as a fairly strong admixture of 6p and 5d orbitals of j = 1/2 or 3/2, as one can expect a much smaller spin–orbit splitting in the axial field of the molecule (about 1/3 of the atomic spin–orbit splitting for the P state). This picture also finds experimental support in recent measurements of hyperfine constants (d and eq0Q) for the ground and Π1/2 excited state of YbF9, where a simple ligand-field model disregarding the contributions from 5dπ orbitals predicted values of d a factor of 2 larger than the measurements. For bond distances much larger that the equilibrium one the system gets closer to the configurations in Yb+ with a dominating 5d contribution.
As already mentioned this energy range above 18000 cm−1 is dense with a large number of excited states that can mix with each other and result in new mixed states, like the [557] and [561] ones.24 These will be addressed in Section 3.6.
Uttam et al.27 reported three unidentified states with energies above 22000 cm−1 which are listed in Table 7 and cannot uniquely be identified with the current results. The one at 26014.8 cm−1 has a larger vibrational constant indicating a more strongly bound state, possibly of the 4f13σ26s configuration. The vibrational spacing of the two states at 23
000 cm−1 rather points to states with a closed f shell.
In the following we only consider the IHFS-CCSD potential energy curves, as the spectroscopic parameters are more reliable for CCSD than for KRCI. The coupled cluster results for the two methods are quite similar, and FS-CCSD was selected (because it does not include the (2h,1p) and (1h,2p) transitions with rather large uncertainties). Fig. 3 contains the original FS-CCSD curves as well as the ones obtained after adiabatzation with three different coupling constants. Looking at the potential energy curves for this energetic region, there are two Ω = 1/2 and two Ω = 3/2 states of Yb(4f146p1)F and Yb(4f145d1)F configurations originating from the (0h,1p) sector. For both Ω values there is an additional state with a configuration stemming from the (1h,0p) sector. By looking at the KRCI results one expects additional states belonging to the
configuration for this energy range, which will not be included in the current considerations.
![]() | ||
Fig. 3 Frank-Condon factors before and after adiabatization for the IHFS-CCSD potential energy curves. C is the coupling strength in Hartree. The lowest 10 vibrational levels of the ground state as well as the lowest 60 vibrational levels of the excited state were computed using the LEVEL program.82 The experimental values18,24,26 have been added as straight lines, the labels are defined in the introduction and Table 7. |
As already mentioned earlier the lowest Ω = 1/2 and Ω = 3/2 states can be identified clearly and assigned to experimental observations. There are several experimental states in this energy region attributed to the mixing of states. The [557] and [561] ones24 are assumed to arise from a mixing of the 31/2 and 41/2. The vibrational constant of the perturbing state (41/2) was estimated to be 605 cm−1 in ref. 18. This agrees with the 41/2 state in Fig. 3 with a configuration, see also Table 7. [574] and [578]26 have not been identified and since their Ω value is unknown, we were not able to assign them to a configuration.
Next we take a look at the changes introduced by adiabatization. For small and intermediate coupling strengths there are no major differences in the potential energy curves, although close to the crossing points the potentials are deformed. Intermediate coupling strengths with slightly deformed potentials close to the avoided crossings will be the most realistic description. For very large coupling strengths one obtains parallel potential energy curves due to the strong repulsion. This also results in a major change of the spectra above 19000 cm−1. One of the differences between the adiabatic spectrum and the upermost one in Fig. 3 is that the Frank-Condon factors of the 41/2 state, which is of the Yb(4f13σ26s)F configuration, are now noticeable and the spacing of the energy levels of the 51/2 is changed. Similarly, transitions belonging to the 51/2 appear.
The influence of adiabatization on spectroscopic parameters can be investigated by comparing spectroscopic constants calculated for the IHFS-CCSD curves without and with a coupling of 100 cm−1 (Table 8). We observe that for this coupling strength, there are small but non-negligible changes for the excitation energies, harmonic frequencies and anharmonicity constants, for all but the fourth and fifth Ω = 1/2 states; there, the coupling does seem to significantly change the anharmonicity constants. Equilibrium distances, on the other hand, are largely unperturbed in all cases. Furthermore, as expected from the preceding discussion, no changes are observed for the ground-state, since it is too separated in energy from the other electronic states.
Ω | CBS | C = 100 cm−1 | ||||||||
---|---|---|---|---|---|---|---|---|---|---|
State | r e | ω e | ω e χ e | T e | State | r e | ω e | ω e χ e | T e | |
1/2 | f14 – 1 | 2.018 | 515 | 2.9 | 0 | 1 | 2.018 | 515 | 2.8 | 0 |
f13 – 2 | 1.940 | 599 | 2.8 | 9627 | 2 | 1.940 | 599 | 2.8 | 9617 | |
f14 – 2 | 1.995 | 539 | 2.6 | 18![]() |
3 | 1.995 | 538 | 2.6 | 18![]() |
|
f13 – 3 | 1.940 | 597 | 2.8 | 20![]() |
4 | 1.935 | 603 | 8.6 | 20![]() |
|
f14 – 3 | 2.003 | 533 | 2.7 | 21![]() |
5 | 2.002 | 586 | 0.4 | 21![]() |
|
f14 – 4 | 1.964 | 581 | 1.8 | 31![]() |
6 | 1.964 | 581 | 1.8 | 31![]() |
|
3/2 | f13 – 1 | 1.944 | 595 | 2.8 | 10![]() |
1 | 1.944 | 594 | 2.8 | 10![]() |
f14 – 1 | 1.992 | 542 | 2.6 | 19![]() |
2 | 1.992 | 542 | 2.6 | 19![]() |
|
f13 – 2 | 1.948 | 591 | 2.8 | 21![]() |
3 | 1.948 | 591 | 2.8 | 21![]() |
|
f14 – 2 | 2.012 | 512 | 2.7 | 24![]() |
4 | 2.012 | 512 | 2.7 | 24![]() |
|
5/2 | f13 – 1 | 1.949 | 589 | 2.8 | 10![]() |
1 | 1.949 | 589 | 2.8 | 10![]() |
f13 – 2 | 1.950 | 584 | 2.8 | 22![]() |
2 | 1.950 | 583 | 2.8 | 22![]() |
|
f14 – 1 | 2.009 | 515 | 2.8 | 24![]() |
3 | 2.009 | 516 | 2.7 | 24![]() |
In order to achieve such a balanced description, our strategy consisted of starting from YbF+ and YbF−, in order to arrive at the wavefunctions for YbF through the (1h,0p) and (0h,1p) sectors of Fock space. Once obtained, electronic states with same Ω values coming from these different sectors are further coupled through a simple adiabatization model in which the coupling strength is taken as a constant.
As a general rule we find that the CI calculations do capture the essential physics of the system, though they are not as reliable as the coupled cluster approaches for excitation energies, bond lengths, harmonic vibrational frequencies and anharmonic constants. In effect, the coupled cluster calculations for the (1h,0p) and (0h,1p) sectors yield the same potential energy curves for the ground state, for internuclear distances up to around 2.8 Å, which is sufficient to capture the bound regions of all states under consideration,
We have determined that the lowest lying excited states arise from the configuration, with transition energies of around 10
000 cm−1, and a splitting about 2000 cm−1. These states are, however, not generally accessible in experiment due to their low dipolar intensity and significantly shifted minima of the potential energy curve resulting in small Frank-Condon factors.
The next set of states, coming above 18000 cm−1, arise from the Yb(4f146p1)F, Yb(4f145d1)F,
and
configurations. Among these, the
configurations generally display the shortest equilibrium distances and deepest potential well, while the Yb(4f145d1)F and Yb(4f146p1)F configurations exhibits the largest bond distances and smallest harmonic frequencies, with the other configurations falling somewhere in between. The lowest Ω = 1/2 and Ω = 3/2 states of this group show a Yb(4f146p1)F orbital composition around the ground-state equilibrium structure, though for longer bond lengths they asymptotically approach the Yb(4f145d1)F configuration.
We note that configurations with three unpaired electrons, such as were only considered with the KRCI method, which has larger uncertainties. This only allows us to make some qualitative statements, e.g. that their bond distances and vibrational constant should be between the values for the other configurations and that they should be higher in energy than the lowes excited Yb(4f146p1)F and
states.
A simple method was applied in order to adiabatize the curves obtained for different sectors and reference wave functions. It was applied to potential energy curves between 18000 and 26
000 cm−1 and small changes of the Franck–Condon factors were observed. The influence on spectroscopic constant was minor, with the exception of the asymmetry constant for two states. However, the approximation introduced (same coupling strength for all states and all geometries) is perhaps not flexible enough, and more sophisticated models should be investigated.
Footnote |
† Electronic supplementary information (ESI) available: A supplementary pdf file is available containing orbitals, additional information about dipole moments and additional Franck–Condon factors. See DOI: 10.1039/d1cp03701c |
This journal is © the Owner Societies 2021 |