Sun et al. [J. Chem. Phys. 144, 191101 (2016)] suggested that common density-functional approximations (DFAs) should exhibit large energy errors for excited states as a necessary consequence of orbital nodality. Motivated by self-interaction corrected density-functional calculations on many-electron systems, we continue their study with the exactly solvable 1s, 2p, and 3d states of 36 hydrogenic one-electron ions (H–Kr35+) and demonstrate with self-consistent calculations that state-of-the-art DFAs indeed exhibit large errors for the 2p and 3d excited states. We consider 56 functionals at the local density approximation (LDA), generalized gradient approximation (GGA) as well as meta-GGA levels, and several hybrid functionals such as the recently proposed machine-learned DM21 local hybrid functional. The best non-hybrid functional for the 1s ground state is revTPSS. As predicted by Sun et al., the 2p and 3d excited states are more difficult for DFAs, and LDA functionals turn out to yield the most systematic accuracy for these states among non-hybrid functionals. The best performance for the three states overall is observed with the BHandH global hybrid GGA functional, which contains 50% Hartree–Fock exchange and 50% LDA exchange. The performance of DM21 is found to be inconsistent, yielding good accuracy for some states and systems and poor accuracy for others. Based on these results, we recommend including a variety of one-electron cations in future training of machine-learned density functionals.

Density-functional theory1,2 (DFT) has become one of the workhorses of computational chemistry, material science, and related fields, as modern density-functional approximations (DFAs) only require a reasonable amount of computational effort while providing a level of accuracy sufficient for semi-quantitative predictions on a broad range of systems.3–5 Although hundreds of DFAs have been proposed, thus forming the infamous zoo of DFAs,6 new DFAs continue to be developed with the aim to find more universally applicable DFAs that combine suitable levels of accuracy and numerical effort.

New DFAs can be constructed along various strategies.5,7–9 The traditional route to construct DFAs is to start from first principles and to impose known limits and constraints; this is the way along which many well-known functionals such as PBE,10 TPSS,11,12 and SCAN13 have been constructed.

Semi-empirical fitting is another route for constructing DFAs. Here, the general idea is to introduce flexibility in the functional form by introducing several independent DFA components that are weighted by parameters, which are optimized against some training dataset. Classical examples of semi-empirically fitted functionals include B3LYP,14 Becke’s 1997 functional15 (B97), and several refinements thereof such as the HCTH [Hamprecht–Cohen–Tozer–Handy] functionals by Handy and co-workers,16,17 as well as the Minnesota family of DFAs by Truhlar and co-workers18,19 that has been reviewed by Mardirossian and Head-Gordon.20 

In reality, the classification of functionals into ones built solely from first principles vs ones formed by semi-empirical fitting is not always clear: for instance, the TPSS exchange functional is parameterized to yield the exact energy for the hydrogen atom’s exact ground state density,11,12 while the SCAN functional13 includes parameters that are fit to data on noble gases. Modern semi-empirical functionals,21–26 in turn, typically employ a combination of the two approaches by restricting the fits to known constraints.

DFAs from either route are widely used, given their suitable numerical accuracy and reasonable computational effort. However, the functionals obtained from the two routes tend to exhibit different behavior. For instance, while semi-empirical DFAs often deliver excellent descriptions of the total energy, they may fail to reproduce electronic densities of the same quality: a famous article by Medvedev et al.27 initiated an intense debate about this in the literature;27–32 it was even pointed out that any general mathematical measure of density error is too arbitrary to be universally useful.33 DFAs built on physical first principles, in contrast, often yield steady performance in a variety of applications, but may not achieve the same level of accuracy as tailored functionals for specific types of systems.

One of the most important limitations of present-day DFAs, regardless of their design, is the self-interaction error (SIE): an artificial interaction of the electrons with themselves. This error is related to density delocalization error and the fractional electron problem,34,35 and leads to incorrect dissociation limits36 and barrier heights,37 for instance. Recent avenues for circumventing SIE in DFAs involve determining the electron density with another method, such as Hartree–Fock38,39 or multiconfigurational wave function theory.40,41 Other types of approaches have also been proposed in the literature. To solve the self-interaction problem, Perdew and Zunger42 (PZ) proposed an orbital-by-orbital self-interaction correction (SIC),

(1)

where EKS is the Kohn–Sham (KS) energy functional,2 and the self-interaction error (SIE) is defined by

(2)

Here, n is the electron density of the i-th occupied orbital with spin σ, and EJ and Exc denote the Coulomb and exchange-correlation energy functionals, respectively. The idea behind PZ-SIC is that the self-interaction error defined by Eq. (2) vanishes for the exact functional,42 and thereby, the Perdew–Zunger functional of Eq. (1) is a better estimate for the total energy than the uncorrected Kohn–Sham DFA EKS; indeed, the PZ functional is exact for one-electron systems such as the H2+ molecule with approximate DFAs.

Despite the simple logic used to construct the PZ-SIC functional, the PZ-SIC method turns out to be quite complicated. The introduction of the explicit orbital dependence in Eqs. (1) and (2) breaks the unitary invariance of the energy functional,43 requiring costly unitary optimization of the orbitals (see Ref. 44 for discussion). Even though the resulting method is known to correct charge transfer errors and barrier heights, it does not lead to improved atomization energies with GGA and meta-GGA functionals in general.45 

Continued research has illuminated other important theoretical aspects of PZ-SIC. First, the orbital dependence in Eqs. (1) and (2) has been recently shown to require the use of complex-valued orbitals for proper minimization, as real-valued orbitals can be shown to correspond to high-order saddle points.46 When complex-valued orbitals are employed, the total energy is lowered, and PZ-SIC does lead to improved atomization energies for some GGA functionals; however, more accurate atomization energies can be obtained at significantly smaller cost with several standard DFAs.47 

Second, the orbital dependence in Eqs. (1) and (2) has also been shown to lead to the existence of several local minima in the orbital space.46 This problem has been recently shown to persist also in a related SIC method48 based on the use of Fermi–Löwdin orbitals (PZFLO-SIC), where various choices for the orbital descriptors lead to distinct local electronic minima.49 The existence of such local minima is a significant and underappreciated aspect of PZ-SIC and PZFLO-SIC calculations, as finding the true ground state may require extensive sampling of the space of the various possible localized electronic configurations or bonding situations.

Despite their theoretical shortcomings, PZ-SIC and PZFLO-SIC have been found useful in many applications,50–52 and we are positive that several of the aforementioned issues in PZ-SIC and PZFLO-SIC can be addressed by developments in the related theories by changing the way the self-interaction correction is applied. One possible way to achieve improved results would be to revisit DFAs based on the requirements of SIC calculations.53 It is known that present-day DFAs yield poor estimates for the noded electron densities that are involved in SIC calculations.54,55 Sun et al.54 demonstrated that the ground and excited state densities of the hydrogen atom (as well as of H2+, see below) lead to large relative errors in the exchange-correlation energy compared to the exact values, but we are not aware of any self-consistent calculations on this issue.

Following the recent discussion in the literature on the accuracy of DFAs on the electron densities of small atoms and ions27–33 and motivated by the obvious connection of one-electron errors (OEEs) to the PZ-SIC and PZFLO-SIC methods, in this work, we will analyze the OEE of various functionals for the 1s ground state as well as the 2p and 3d excited states of hydrogenic ions Z(Z−1)+, whose exact energies are well known to be given in atomic units by

(3)

where Z is the atomic number, nl + 1 is the principal quantum number, and l is the angular momentum.

As mentioned above, calculations of ground and excited states of the hydrogen atom and the 1σg ground state and 1σu excited state of H2+ have been discussed by Sun et al.54 with non-self-consistent electron densities, while the 1s ground states of hydrogenic mononuclear cations as well as the 1σ ground states of hydrogenic diatomic cations have been discussed recently by Lonsdale and Goerigk56 using self-consistent calculations. The novel contribution of this work is to address (highly) excited states with noded electron densities of hydrogenic cations self-consistently. Importantly, similar to the 1s ground state, the 2p and 3d excited states (as well as the analogous higher excited states like 4f) are the lowest states of the corresponding symmetry, and the ground-state Kohn–Sham scheme is applicable to such excited states as well as shown by Gunnarsson and Lundqvist.57 

We pursue thorough density-functional investigations of the 1s, 2p, and 3d states of hydrogenic ions in benchmark-quality Gaussian basis sets, especially suited for this purpose with a selection of 56 popular DFAs, including the recently developed, highly sophisticated machine-learned DeepMind 21 (DM21) local hybrid functional.58 

The layout of this work is as follows. The computational details are presented in Sec. II, and the results are given in Sec. III. A summary of our findings and an outlook for further investigations are given in Sec. IV. Atomic units are used throughout, unless specified otherwise.

We only use free and open-source software (FOSS) in this work, following the philosophy discussed in Ref. 59. PySCF60 is an electronic structure code for all-electron calculations by using Gaussian-type orbitals (GTOs). As we are targeting one-electron states of specific symmetry (s, p, or d states), following Gunnarsson and Lundqvist,57 we truncate the basis set in all calculations to contain functions only of the pursued symmetry: calculations on the 1s/2p/3d state only include the basis functions of the corresponding symmetry (s, p, or d functions, respectively) from the chosen parent basis set. This procedure has two important features: the 2p and 3d excited states become the ground state in the reduced-basis calculation, and the computational requirements are smaller since fewer integrals need to be calculated in the reduced basis than in the original basis set.

The one-electron guess—which is exact for one-electron systems and thereby is also expected to be accurate for calculations employing DFAs as well—is used in all calculations.61 To ensure that the SCF procedure converges to the global minimum instead of a saddle point, the following procedure was used. First, a regular SCF calculation was performed with PySCF with default settings; direct inversion in the iterative subspace (DIIS) is used to accelerate these calculations.62,63 Next, convergence to saddle point solutions was checked: cases where the SCF converged to a final energy higher than that of the initial guess were restarted, with new calculations employing iterative diagonalization with level shifting64 instead of DIIS to converge to the ground state. All calculations reported in this work are fully converged to a threshold of 1 × 10−7Eh.

For the GTO basis sets, we use the family of hydrogenic Gaussian basis sets65 (HGBS-n) that have been designed for high-accuracy calculations on atoms and small molecules. A special feature of the HGBS basis sets is that the basis for atomic number Z is determined by a universal even-tempered basis set for the ions Y(Y−1)+ for Y ∈ [1, Z], whereas augmented hydrogenic Gaussian basis sets (AHGBS-n) add further functions for describing the Z = 0.5 one-electron ion.65 The parameter n controls the relative precision of the hydrogenic Gaussian basis; (A)HGBS-n reproduces the exact total energies of the one-electron ions to an approximate relative accuracy of 10n.65 The motivation of this approach in Ref. 65 was that a many-electron atom experiences a screened nuclear charge that can be rewritten in terms of a radially dependent effective charge Zeff = Zeff(r) with the asymptotic limits Zeff(0) = Z and either Zeff() = Z with the asymptotic limit Z = 0 for Hartree–Fock and DFT or Z = 1 for the exact effective potential.61 

Another feature of the HGBS basis sets is that the functions of various angular symmetries are determined independently of each other, which facilitates the formation of polarized counterparts of the basis sets that are essential for studying molecules and excited states, as additional shells are added to the basis like lego blocks.65 Following Ref. 65, the basis set with p ≥ 1 polarization shells and accuracy n is denoted as (A)HGBSPp-n. The definition of polarization shells varies by atom (see Ref. 65 for discussion); however, as we only include the functions of the pursued symmetry in each calculation, the choice of the polarization level of the (A)HGBSPp-n basis set does not matter as long as the original basis contains functions of the highest targeted angular momentum for the targeted atom, that is, d functions in this work.

For the reasons listed above, the hydrogenic Gaussian basis sets of Ref. 65 are ideally suited for the present study—as will be demonstrated in Sec. III by benchmarks with functions from the polarization consistent (pc-n) basis sets66 and their augmented versions67 (aug-pc-n)—and, as will be discussed in Sec. III A, we will take the exponents from the AHGBSP3-n basis sets in this work. All basis sets were taken from the Basis Set Exchange.68 

The Libxc library9—which implements over 600 DFAs—is used in PySCF to evaluate the DFAs. The library provides access to a vast variety of DFAs, of which 49 were chosen for this work; see Table I for the complete list of investigated functionals. Our selection includes functionals from the first to the fourth rung of Jacob’s ladder,123 that is, local density approximations (LDAs), generalized gradient approximations (GGAs), meta-GGAs, as well as global and range-separated hybrid functionals. In addition, we consider six hybrids of rSCAN and r2SCAN with varying fractions of Hartree–Fock exchange discussed in Ref. 120; these functionals were defined in the PySCF input files as weighted combinations of r(2)SCAN exchange and Hartree–Fock exchange + 100% r(2)SCAN correlation. The DM21 functional was also chosen for this study; we use the original implementation in PySCF of Kirkpatrick et al.58 This brings up the total to 56 functionals for this study. An unpruned (300,590) quadrature grid is used in all calculations, including the non-local correlation component in B97M-V, ωB97M-V, and LC-VV10.

TABLE I.

List of investigated functionals, including the publication year, the Libxc identifier, the calculated MSEs for the 1s, 2p, and 3d states as well as the respective OE. Tables containing functional rankings by error for the individual states as well as the OE can be found in the supplementary material (Tables S1–S4). The Libxc identifiers contain information about the functional; in addition to the rung of Jacob’s ladder: LDA, GGA, or meta-GGA (mGGA), hybrid (hyb) functionals are also identifiable from the list.

MSE
NameYearLibxc identifier1s2p3dOE
ωB97M-V23  2016 HYB_MGGA_XC_WB97M_V 8.106 × 10−4 5.987 × 10−3 1.128 × 10−2 6.027 × 10−3 
ωB97X-D69  2008 HYB_GGA_XC_WB97X_D 7.641 × 10−4 1.385 × 10−2 2.512 × 10−2 1.324 × 10−2 
B3LYP14,71–75 1994 HYB_GGA_XC_B3LYP 7.203 × 10−4 1.055 × 10−2 2.464 × 10−2 1.197 × 10−2 
B97-116  1998 HYB_GGA_XC_B97_1 3.756 × 10−4 1.108 × 10−2 2.436 × 10−2 1.194 × 10−2 
B97M-V22  2015 MGGA_XC_B97M_V 2.466 × 10−4 5.258 × 10−3 1.491 × 10−2 6.804 × 10−3 
BHandH73  1993 HYB_GGA_XC_BHANDH 5.122 × 10−3 2.822 × 10−3 4.341 × 10−4 2.793 × 10−3 
BLOC74,75 2013 MGGA_X_BLOC, MGGA_C_REVTPSS 2.046 × 10−5 1.094 × 10−2 2.320 × 10−2 1.139 × 10−2 
BLYP71–72  1988 GGA_X_B88, GGA_C_LYP 5.790 × 10−4 1.209 × 10−2 2.857 × 10−2 1.374 × 10−2 
BLYP3576,77 2011 HYB_GGA_XC_BLYP35 3.902 × 10−4 7.824 × 10−3 1.837 × 10−2 8.861 × 10−3 
BOP70,78 1999 GGA_X_B88, GGA_C_OP_B88 5.789 × 10−4 1.209 × 10−2 2.857 × 10−2 1.374 × 10−2 
CAM-B3LYP79  2004 HYB_GGA_XC_CAM_B3LYP 8.286 × 10−4 7.569 × 10−3 1.617 × 10−2 8.188 × 10−3 
CAM-QTP0080  2014 HYB_GGA_XC_CAM_QTP_00 4.572 × 10−4 4.187 × 10−3 8.300 × 10−3 4.315 × 10−3 
CAM-QTP0181  2016 HYB_GGA_XC_CAM_QTP_01 1.339 × 10−3 4.961 × 10−3 9.542 × 10−3 5.281 × 10−3 
CAM-QTP0282  2018 HYB_GGA_XC_CAM_QTP_02 1.714 × 10−3 3.302 × 10−3 6.680 × 10−3 3.899 × 10−3 
CHACHIYO83,84 2015 LDA_X,LDA_C_CHACHIYO 7.711 × 10−3 3.476 × 10−3 1.155 × 10−2 7.579 × 10−3 
DM2158  2021 Uses PySCF implementation 2.126 × 10−3 5.435 × 10−3 1.226 × 10−2 6.606 × 10−3 
instead of Libxc 
GAM85  2015 GGA_X_GAM, GGA_C_GAM 2.584 × 10−3 1.561 × 10−2 3.650 × 10−2 1.823 × 10−2 
HCTH-9316  1998 GGA_XC_HCTH_93 9.252 × 10−4 1.642 × 10−2 3.689 × 10−2 1.808 × 10−2 
HSE0386,87 2003 HYB_GGA_XC_HSE03 1.116 × 10−3 1.150 × 10−2 2.453 × 10−2 1.238 × 10−2 
HSE0691–88  2006 HYB_GGA_XC_HSE06 6.659 × 10−4 8.785 × 10−3 2.052 × 10−2 9.989 × 10−3 
HSE1289  2012 HYB_GGA_XC_HSE12 6.187 × 10−4 8.201 × 10−3 1.911 × 10−2 9.309 × 10−3 
LC-QTP82  2018 HYB_GGA_XC_LC_QTP 2.276 × 10−3 3.818 × 10−3 7.590 × 10−3 4.561 × 10−3 
LC-VV1090  2010 HYB_GGA_XC_LC_VV10 1.089 × 10−3 6.905 × 10−3 1.166 × 10−2 6.553 × 10−3 
LRC-ωPBE91  2009 HYB_GGA_XC_LRC_WPBE 1.121 × 10−3 9.294 × 10−3 1.668 × 10−2 9.033 × 10−3 
M06-L92,93 2006 MGGA_X_M06_L, MGGA_C_M06_L 9.242 × 10−4 1.521 × 10−2 3.368 × 10−2 1.660 × 10−2 
M11-L94  2012 MGGA_X_M11_L, MGGA_C_M11_L 2.320 × 10−3 1.876 × 10−2 4.593 × 10−2 2.234 × 10−2 
MN12-L95  2012 MGGA_X_MN12_L, MGGA_C_MN12_L 1.541 × 10−3 5.796 × 10−3 2.036 × 10−2 9.234 × 10−3 
MN1596  2016 HYB_MGGA_X_MN15, 2.532 × 10−4 9.867 × 10−3 2.133 × 10−2 1.048 × 10−2 
MGGA_C_MN15 
MN15-L97  2016 MGGA_X_MN15_L, MGGA_C_MN15_L 2.118 × 10−3 4.673 × 10−3 1.327 × 10−2 6.689 × 10−3 
MS0103–100  2012 MGGA_X_MS0, GGA_C_REGTPSS 5.973 × 10−4 1.121 × 10−2 2.233 × 10−2 1.138 × 10−2 
MS1104–101  2013 MGGA_X_MS1, GGA_C_REGTPSS 6.013 × 10−4 1.161 × 10−2 2.346 × 10−2 1.189 × 10−2 
MS2104–101  2013 MGGA_X_MS2, GGA_C_REGTPSS 6.039 × 10−4 1.193 × 10−2 2.432 × 10−2 1.229 × 10−2 
OLYP71,102 2009 GGA_X_OPTX, GGA_C_LYP 3.934 × 10−4 1.309 × 10−2 2.971 × 10−2 1.440 × 10−2 
PBE10,103 1996 GGA_X_PBE, GGA_C_PBE 9.196 × 10−4 1.096 × 10−2 2.538 × 10−2 1.242 × 10−2 
PBEsol104  2007 GGA_X_PBE_SOL, GGA_C_PBE_SOL 3.799 × 10−3 6.918 × 10−3 2.010 × 10−2 1.027 × 10−2 
PKZB105  1999 MGGA_X_PKZB, MGGA_C_PKZB 9.408 × 10−4 1.038 × 10−2 2.470 × 10−2 1.201 × 10−2 
PW91112–108  1992 GGA_X_PW91, GGA_C_PW91 7.771 × 10−4 1.086 × 10−2 2.449 × 10−2 1.204 × 10−2 
QTP17109  2018 HYB_GGA_XC_QTP17 3.060 × 10−3 1.483 × 10−3 4.612 × 10−3 3.051 × 10−3 
RPBE10,103,110 1999 GGA_X_RPBE, GGA_C_PBE 4.576 × 10−4 1.346 × 10−2 2.932 × 10−2 1.441 × 10−2 
SPW9283,111 1992 LDA_X,LDA_C_PW_MOD 7.702 × 10−3 3.392 × 10−3 1.138 × 10−2 7.490 × 10−3 
SVWN83,112,113 1980 LDA_X, LDA_C_VWN 7.707 × 10−3 3.390 × 10−3 1.138 × 10−2 7.492 × 10−3 
TASK111,114 2019 MGGA_X_TASK, LDA_C_PW 2.523 × 10−3 1.463 × 10−2 2.698 × 10−2 1.471 × 10−2 
TM115  2016 MGGA_X_TM, MGGA_C_TM 2.972 × 10−5 1.035 × 10−2 2.240 × 10−2 1.093 × 10−2 
TPSS11,12 2003 MGGA_X_TPSS, MGGA_C_TPSS 2.046 × 10−5 1.094 × 10−2 2.318 × 10−2 1.138 × 10−2 
TPSSh116  2003 HYB_MGGA_XC_TPSSH 1.640 × 10−5 9.835 × 10−3 2.080 × 10−2 1.022 × 10−2 
XLYP117  2004 GGA_XC_XLYP 1.369 × 10−4 1.217 × 10−2 2.792 × 10−2 1.341 × 10−2 
r2SCAN118,119 2020 MGGA_X_R2SCAN, MGGA_C_R2SCAN 1.495 × 10−5 8.087 × 10−3 1.617 × 10−2 8.091 × 10−3 
r2SCAN0120  2022 Custom-defined in PySCF 8.173 × 10−6 6.051 × 10−3 1.204 × 10−2 6.035 × 10−3 
r2SCAN50120  2022 Custom-defined in PySCF 3.504 × 10−6 4.025 × 10−3 7.975 × 10−3 4.001 × 10−3 
r2SCANh120  2022 Custom-defined in PySCF 1.198 × 10−5 7.271 × 10−3 1.451 × 10−2 7.266 × 10−3 
rSCAN121  2019 MGGA_X_RSCAN, MGGA_C_RSCAN 1.495 × 10−5 8.087 × 10−3 1.617 × 10−2 8.091 × 10−3 
rSCAN0120  2022 Custom-defined in PySCF 8.173 × 10−6 6.051 × 10−3 1.204 × 10−2 6.034 × 10−3 
rSCAN50120  2022 Custom-defined in PySCF 3.504 × 10−6 4.025 × 10−3 7.975 × 10−3 4.001 × 10−3 
rSCANh120  2022 Custom-defined in PySCF 1.198 × 10−5 7.271 × 10−3 1.451 × 10−2 7.266 × 10−3 
revPBE10,103,122 1998 GGA_X_PBE_R, GGA_C_PBE 4.572 × 10−4 1.358 × 10−2 2.995 × 10−2 1.466 × 10−2 
revTPSS99,100 2009 MGGA_X_REVTPSS, 1.302 × 10−5 1.040 × 10−2 2.186 × 10−2 1.076 × 10−2 
MGGA_C_REVTPSS 
MSE
NameYearLibxc identifier1s2p3dOE
ωB97M-V23  2016 HYB_MGGA_XC_WB97M_V 8.106 × 10−4 5.987 × 10−3 1.128 × 10−2 6.027 × 10−3 
ωB97X-D69  2008 HYB_GGA_XC_WB97X_D 7.641 × 10−4 1.385 × 10−2 2.512 × 10−2 1.324 × 10−2 
B3LYP14,71–75 1994 HYB_GGA_XC_B3LYP 7.203 × 10−4 1.055 × 10−2 2.464 × 10−2 1.197 × 10−2 
B97-116  1998 HYB_GGA_XC_B97_1 3.756 × 10−4 1.108 × 10−2 2.436 × 10−2 1.194 × 10−2 
B97M-V22  2015 MGGA_XC_B97M_V 2.466 × 10−4 5.258 × 10−3 1.491 × 10−2 6.804 × 10−3 
BHandH73  1993 HYB_GGA_XC_BHANDH 5.122 × 10−3 2.822 × 10−3 4.341 × 10−4 2.793 × 10−3 
BLOC74,75 2013 MGGA_X_BLOC, MGGA_C_REVTPSS 2.046 × 10−5 1.094 × 10−2 2.320 × 10−2 1.139 × 10−2 
BLYP71–72  1988 GGA_X_B88, GGA_C_LYP 5.790 × 10−4 1.209 × 10−2 2.857 × 10−2 1.374 × 10−2 
BLYP3576,77 2011 HYB_GGA_XC_BLYP35 3.902 × 10−4 7.824 × 10−3 1.837 × 10−2 8.861 × 10−3 
BOP70,78 1999 GGA_X_B88, GGA_C_OP_B88 5.789 × 10−4 1.209 × 10−2 2.857 × 10−2 1.374 × 10−2 
CAM-B3LYP79  2004 HYB_GGA_XC_CAM_B3LYP 8.286 × 10−4 7.569 × 10−3 1.617 × 10−2 8.188 × 10−3 
CAM-QTP0080  2014 HYB_GGA_XC_CAM_QTP_00 4.572 × 10−4 4.187 × 10−3 8.300 × 10−3 4.315 × 10−3 
CAM-QTP0181  2016 HYB_GGA_XC_CAM_QTP_01 1.339 × 10−3 4.961 × 10−3 9.542 × 10−3 5.281 × 10−3 
CAM-QTP0282  2018 HYB_GGA_XC_CAM_QTP_02 1.714 × 10−3 3.302 × 10−3 6.680 × 10−3 3.899 × 10−3 
CHACHIYO83,84 2015 LDA_X,LDA_C_CHACHIYO 7.711 × 10−3 3.476 × 10−3 1.155 × 10−2 7.579 × 10−3 
DM2158  2021 Uses PySCF implementation 2.126 × 10−3 5.435 × 10−3 1.226 × 10−2 6.606 × 10−3 
instead of Libxc 
GAM85  2015 GGA_X_GAM, GGA_C_GAM 2.584 × 10−3 1.561 × 10−2 3.650 × 10−2 1.823 × 10−2 
HCTH-9316  1998 GGA_XC_HCTH_93 9.252 × 10−4 1.642 × 10−2 3.689 × 10−2 1.808 × 10−2 
HSE0386,87 2003 HYB_GGA_XC_HSE03 1.116 × 10−3 1.150 × 10−2 2.453 × 10−2 1.238 × 10−2 
HSE0691–88  2006 HYB_GGA_XC_HSE06 6.659 × 10−4 8.785 × 10−3 2.052 × 10−2 9.989 × 10−3 
HSE1289  2012 HYB_GGA_XC_HSE12 6.187 × 10−4 8.201 × 10−3 1.911 × 10−2 9.309 × 10−3 
LC-QTP82  2018 HYB_GGA_XC_LC_QTP 2.276 × 10−3 3.818 × 10−3 7.590 × 10−3 4.561 × 10−3 
LC-VV1090  2010 HYB_GGA_XC_LC_VV10 1.089 × 10−3 6.905 × 10−3 1.166 × 10−2 6.553 × 10−3 
LRC-ωPBE91  2009 HYB_GGA_XC_LRC_WPBE 1.121 × 10−3 9.294 × 10−3 1.668 × 10−2 9.033 × 10−3 
M06-L92,93 2006 MGGA_X_M06_L, MGGA_C_M06_L 9.242 × 10−4 1.521 × 10−2 3.368 × 10−2 1.660 × 10−2 
M11-L94  2012 MGGA_X_M11_L, MGGA_C_M11_L 2.320 × 10−3 1.876 × 10−2 4.593 × 10−2 2.234 × 10−2 
MN12-L95  2012 MGGA_X_MN12_L, MGGA_C_MN12_L 1.541 × 10−3 5.796 × 10−3 2.036 × 10−2 9.234 × 10−3 
MN1596  2016 HYB_MGGA_X_MN15, 2.532 × 10−4 9.867 × 10−3 2.133 × 10−2 1.048 × 10−2 
MGGA_C_MN15 
MN15-L97  2016 MGGA_X_MN15_L, MGGA_C_MN15_L 2.118 × 10−3 4.673 × 10−3 1.327 × 10−2 6.689 × 10−3 
MS0103–100  2012 MGGA_X_MS0, GGA_C_REGTPSS 5.973 × 10−4 1.121 × 10−2 2.233 × 10−2 1.138 × 10−2 
MS1104–101  2013 MGGA_X_MS1, GGA_C_REGTPSS 6.013 × 10−4 1.161 × 10−2 2.346 × 10−2 1.189 × 10−2 
MS2104–101  2013 MGGA_X_MS2, GGA_C_REGTPSS 6.039 × 10−4 1.193 × 10−2 2.432 × 10−2 1.229 × 10−2 
OLYP71,102 2009 GGA_X_OPTX, GGA_C_LYP 3.934 × 10−4 1.309 × 10−2 2.971 × 10−2 1.440 × 10−2 
PBE10,103 1996 GGA_X_PBE, GGA_C_PBE 9.196 × 10−4 1.096 × 10−2 2.538 × 10−2 1.242 × 10−2 
PBEsol104  2007 GGA_X_PBE_SOL, GGA_C_PBE_SOL 3.799 × 10−3 6.918 × 10−3 2.010 × 10−2 1.027 × 10−2 
PKZB105  1999 MGGA_X_PKZB, MGGA_C_PKZB 9.408 × 10−4 1.038 × 10−2 2.470 × 10−2 1.201 × 10−2 
PW91112–108  1992 GGA_X_PW91, GGA_C_PW91 7.771 × 10−4 1.086 × 10−2 2.449 × 10−2 1.204 × 10−2 
QTP17109  2018 HYB_GGA_XC_QTP17 3.060 × 10−3 1.483 × 10−3 4.612 × 10−3 3.051 × 10−3 
RPBE10,103,110 1999 GGA_X_RPBE, GGA_C_PBE 4.576 × 10−4 1.346 × 10−2 2.932 × 10−2 1.441 × 10−2 
SPW9283,111 1992 LDA_X,LDA_C_PW_MOD 7.702 × 10−3 3.392 × 10−3 1.138 × 10−2 7.490 × 10−3 
SVWN83,112,113 1980 LDA_X, LDA_C_VWN 7.707 × 10−3 3.390 × 10−3 1.138 × 10−2 7.492 × 10−3 
TASK111,114 2019 MGGA_X_TASK, LDA_C_PW 2.523 × 10−3 1.463 × 10−2 2.698 × 10−2 1.471 × 10−2 
TM115  2016 MGGA_X_TM, MGGA_C_TM 2.972 × 10−5 1.035 × 10−2 2.240 × 10−2 1.093 × 10−2 
TPSS11,12 2003 MGGA_X_TPSS, MGGA_C_TPSS 2.046 × 10−5 1.094 × 10−2 2.318 × 10−2 1.138 × 10−2 
TPSSh116  2003 HYB_MGGA_XC_TPSSH 1.640 × 10−5 9.835 × 10−3 2.080 × 10−2 1.022 × 10−2 
XLYP117  2004 GGA_XC_XLYP 1.369 × 10−4 1.217 × 10−2 2.792 × 10−2 1.341 × 10−2 
r2SCAN118,119 2020 MGGA_X_R2SCAN, MGGA_C_R2SCAN 1.495 × 10−5 8.087 × 10−3 1.617 × 10−2 8.091 × 10−3 
r2SCAN0120  2022 Custom-defined in PySCF 8.173 × 10−6 6.051 × 10−3 1.204 × 10−2 6.035 × 10−3 
r2SCAN50120  2022 Custom-defined in PySCF 3.504 × 10−6 4.025 × 10−3 7.975 × 10−3 4.001 × 10−3 
r2SCANh120  2022 Custom-defined in PySCF 1.198 × 10−5 7.271 × 10−3 1.451 × 10−2 7.266 × 10−3 
rSCAN121  2019 MGGA_X_RSCAN, MGGA_C_RSCAN 1.495 × 10−5 8.087 × 10−3 1.617 × 10−2 8.091 × 10−3 
rSCAN0120  2022 Custom-defined in PySCF 8.173 × 10−6 6.051 × 10−3 1.204 × 10−2 6.034 × 10−3 
rSCAN50120  2022 Custom-defined in PySCF 3.504 × 10−6 4.025 × 10−3 7.975 × 10−3 4.001 × 10−3 
rSCANh120  2022 Custom-defined in PySCF 1.198 × 10−5 7.271 × 10−3 1.451 × 10−2 7.266 × 10−3 
revPBE10,103,122 1998 GGA_X_PBE_R, GGA_C_PBE 4.572 × 10−4 1.358 × 10−2 2.995 × 10−2 1.466 × 10−2 
revTPSS99,100 2009 MGGA_X_REVTPSS, 1.302 × 10−5 1.040 × 10−2 2.186 × 10−2 1.076 × 10−2 
MGGA_C_REVTPSS 

As the total energies scale as EnZ2 according to Eq. (3), the results will be analyzed in terms of absolute relative errors (AREs). Hydrogenic estimates show that the approximated exchange-correlation energy scales like Z in the large Z limit,124 meaning that the relative errors should tend to zero like 1/Z. The ARE for a given state of a given ion is given by

(4)

The information in the AREs is analyzed with two more error metrics. The mean state error (MSE) measures the overall functional error over all ions by averaging the ARE over all ions,

(5)

The overall error (OE) for a functional is obtained by further averaging the MSE over all considered states (1s, 2p, and 3d),

(6)

Before pursuing density-functional calculations, we analyze the basis set truncation errors (BSTEs) for the one-electron cations in the polarization consistent and hydrogenic Gaussian basis sets. We aim for a mean BSTE smaller than 5 × 10−5Eh for the whole benchmark set ranging from H0 to Kr35+ to ensure that our results are converged close to the complete basis set limit.

Unrestricted Hartree–Fock (UHF) is exact for one-electron systems and thereby gives the exact energy EnUHF in the studied basis; therefore, the difference between EnUHF/basis and the exact analytical energy [Eq. (3)],

(7)

is a variational measure of the BSTE for the state with given n of the studied hydrogenic ions.

The calculated mean BSTEs for a variety of polarization consistent and hydrogenic Gaussian basis sets are shown in Fig. 1; additional results can be found in the supplementary material. Unsurprisingly, uncontracting the (aug-)pc-n basis sets—yielding the unc-(aug-)pc-n basis sets—results in a noticeable decrease in the BSTE, because the contractions were determined in Ref. 66 with the BLYP functional71–72 that suffers from SIE for the 1s state, while the p and d functions in the basis set describe either polarization effects or the occupied p or d orbitals in the screened neutral atom. Although the large uncontracted polarization consistent basis sets exhibit satisfactory performance for the 1s state, they result in much larger errors for the 2p and 3d states; this error is again caused by the p and d orbitals in the neutral atom being screened by the core electrons, which results in the lack of tight p and d basis functions that are necessary for the 2p and 3d states of the one-electron ions.

FIG. 1.

Mean basis set truncation error (ME) in Eh at UHF level of theory for the 1s ground state and the 2p and 3d excited states, respectively. The reference values are calculated with Eq. (3), and the aimed accuracy threshold 5 × 10−5Eh is shown with the dashed horizontal line.

FIG. 1.

Mean basis set truncation error (ME) in Eh at UHF level of theory for the 1s ground state and the 2p and 3d excited states, respectively. The reference values are calculated with Eq. (3), and the aimed accuracy threshold 5 × 10−5Eh is shown with the dashed horizontal line.

Close modal

In contrast, the primitive (not contracted) hydrogenic Gaussian basis sets of Ref. 65 show uniform accuracy for the 1s, 2p, and 3d states, and as can be observed in Fig. 1, the targeted mean BSTE threshold is roughly achieved already with the AHGBSP3-7 basis set. The AHGBSP3-9 basis sets yield errors below the desired threshold for all states and is, therefore, chosen for all the remaining calculations of this study.

Although this analysis was limited to Hartree–Fock calculations, we note that the basis set requirements of Hartree–Fock and DFT are known to be similar.125 Furthermore, reliable reference energies for DFAs can be obtained with fully numerical methods,137–128 and exploratory calculations presented in the supplementary material confirm that the BSTEs in the AHGBSP3-9 basis are also small for DFAs.

1. Exploratory analysis

We begin the analysis with a graphical study of the results of the SPW92, PBEsol, revTPSS, MN15-L, BHandH, and DM21 functionals in Fig. 2. This limited set of functionals contains LDA, GGA, and meta-GGA functionals from first principles (SPW92, PBEsol, and revTPSS, respectively), semiempirical functionals (MN15-L and DM21), as well as hybrid functionals (BHandH and DM21).

FIG. 2.

Functional errors for the 1s, 2p, and 3d states for the SPW92, PBEsol, revTPSS, MN15-L, BHandH, and DM21 functionals.

FIG. 2.

Functional errors for the 1s, 2p, and 3d states for the SPW92, PBEsol, revTPSS, MN15-L, BHandH, and DM21 functionals.

Close modal

As will be discussed in Sec. III B 2, revTPSS is the most accurate meta-GGA functional for the 1s state. In Fig. 2, revTPSS is outperformed by DM21 only for He+; otherwise, revTPSS affords much lower errors than the five other functionals in the figure. In contrast, the performance of DM21 is inconsistent. DM21 has lower errors for light ions than for heavy ions, but the curve is kinked for the light ions. DM21’s curve becomes smooth for heavy ions, but DM21 is also less accurate for heavy ions. MN15-L also shows a kinky behavior with lower errors for light ions; these non-systematic features of DM21 and MN15-L can be tentatively explained by their semiempirical character; the curves for the first principles functionals are smoother.

The functional errors for the 2p state are shown in Fig. 2(b). The performance for the 2p state is strikingly different compared to the 1s state shown in Fig. 2(a). The plots for the 2p state in Fig. 2(b) show more structure and curve crossings. The behavior of DM21 is qualitatively different from that of the other functionals: DM21 shows large relative errors for light atoms and lower relative errors for heavy atoms, while most of the other functionals shown behave similarly to each other. The only other exceptions to this are the SPW92 and MN15-L functionals that show dips at Z ≃ 3 and Z ≃ 4, respectively; the two functionals are thus oddly more accurate for some values of Z than others.

The errors for the 3d state are shown in Fig. 2(c). BHandH has small errors for all ions for the 3d state. The behavior of DM21 and MN15-L again differs qualitatively from the other functionals. Although DM21 shows less variation for the 3d state than for the 2p state, MN15-L does the opposite: MN15-L has large errors for light ions and becomes nearly as accurate as BHandH for Z ≃ 22; however, the relative error increases again for heavier ions.

2. Full analysis

The MSEs and OE for all studied functionals are shown in Table I. Although Table I contains all of the data used in the present analysis, additional tables showing the rankings of the functionals in terms of the errors for the 1s, 2p, and 3d states as well as in terms of the overall error can be found in the supplementary material.

Clearly, the performance of all LDAs is practically identical. This suggests that the functional error for LDAs is limited by the simple functional form. Although LDAs show larger errors than GGAs and meta-GGAs for the 1s state, they perform better than GGAs and meta-GGAs for the 2p and 3d states.

PBEsol104 is the GGA that yields the smallest errors for the 2p and 3d states, as well as the smallest overall error. The XLYP GGA has a lower error than PBEsol for the 1s state. Although XLYP and even PBEsol are better for 1s states than any LDA, they have higher OEs than any LDA because of their considerably poorer performance for the 2p and 3d states. Analogous findings apply also to all other studied GGAs.

The best meta-GGA for the 1s state is revTPSS,99,100 closely followed by rSCAN121 and r2SCAN118 (see Table I or the supplementary material). The best meta-GGA in terms of overall error is MN15-L.97 

Hybrid functionals have better accuracy as they contain some Hartree–Fock exchange that is free of self-interaction. The best hybrid GGA functionals in terms of overall error are BHandH73 and QTP17.109 BHandH has low MSEs for all states and has the best overall performance, which can be understood by its composition of 50% of Hartree–Fock exchange and 50% LDA exchange + 100% Lee–Yang–Parr correlation. QTP17 has the second-best performance for all states; it, too, contains a mixture of Hartree–Fock (62%) and LDA exchange (38%).

The best functionals of each rung in terms of overall error are SPW92, PBEsol, MN15-L, and BHandH. The corresponding error distributions are summarized in comparison to DM21 in Fig. 3. Interestingly, the performance of the DM21 functional appears similar to that of MN15-L.

FIG. 3.

MSEs and OEs for the best functional of each rung of Jacob’s ladder of all the investigated functionals.

FIG. 3.

MSEs and OEs for the best functional of each rung of Jacob’s ladder of all the investigated functionals.

Close modal

Following Medvedev et al.,27 the calculated OE for all functionals and ions is plotted against the publication year in Fig. 4. As is clear from this plot, the improvement in one-electron error is not fully systematic and features a significant amount of spread and some notable outliers such as M11-L, GAM, and TASK. However, in the recent decade, hybrid functionals have become better overall. As an example, the various QTP functionals dominate the bottom right of the figure; these functionals are closely related in functional form and contain high amounts of Hartree–Fock exchange, which decreases the one-electron error. Unsurprisingly, hybrid functionals based on rSCAN and r2SCAN perform well, and the functionals with large fractions of Hartree–Fock exchange share the bottom right of the figure with the QTP functionals.

FIG. 4.

The overall functional error (OE) for various types of functionals plotted as a function of the publication year. Yearly averages are shown as solid lines as a guide to the eye.

FIG. 4.

The overall functional error (OE) for various types of functionals plotted as a function of the publication year. Yearly averages are shown as solid lines as a guide to the eye.

Close modal

3. Comparison to literature data

The study of Lonsdale and Goerigk56 employed an uncontracted aug-cc-pVQZ basis set129,130 with the following nucleus-dependent quadrature grids for the study of 1s states of hydrogenic cations: (45,770) for H and He, (50,770) for Li–Ne, (55,770) for Na–Ar, and (60,770) for K–Kr. However, it appears that K and Ca were excluded from Ref. 56 (see Figs. 3 and 10 in Ref. 56).

Our trends and absolute values for the MSE for the 1s state are in satisfactory agreement for the subset of functionals studied in both works, although we did identify basis set incompleteness issues in some results of Ref. 56 as discussed in the supplementary material. The largest basis set incompleteness effects are observed for the M11-L and M06-L Minnesota functionals, which are known to converge remarkably slowly to the basis set limit.131 

Lonsdale and Goerigk56 only studied one LDA functional (SVWN); we considered more LDAs and found them to have similar performance. Lonsdale and Goerigk56 found OLYP to be the best GGA functional for the 1s state; we also considered XLYP and found it to yield a considerably lower MSE for the 1s state than OLYP. Lonsdale and Goerigk56 included a broader set of hybrid functionals separating global, range-separated hybrids and double hybrids; however, our main motivation is the connection to self-interaction corrected methods where hybrid functionals are typically not used. We found rSCAN50/r2SCAN50 to be the best hybrid functional for the 1s state, while Lonsdale and Goerigk56 determined TPSSh and SCAN0 to be the best hybrids. All rSCAN- and r2SCAN-based hybrid functionals, i.e., rSCANh/r2SCANh, rSCAN0/r2SCAN0, rSCAN50/r2SCAN50, and TPSSh, have a good performance for the 1s state. The revTPSS functional is found in our work as well by Lonsdale and Goerigk56 to be the best non-hybrid meta-GGA functional for the 1s state.

We used exactly solvable hydrogenic cations in their 1s ground state and 2p and 3d excited states to determine the self-consistent one-electron error for 56 density functionals including the novel DM21 of Kirkpatrick et al.,58 employing the methodology of Gunnarsson and Lundqvist57 for the excited state calculations. In accordance with an earlier finding by Sun et al.54 apparently based on non-self-consistent calculations for the hydrogen atom and molecule and one LDA functional, we find for 36 hydrogenic cations that all LDAs perform better for the excited 2p and 3d states than any of the tested GGAs and meta-GGAs. The performance of various LDAs appears to be almost identical since the calculated errors are nearly indistinguishable, suggesting that the errors are limited by the simple functional form used in LDAs. Sun et al.54 pointed out that larger errors for excited states are a necessary consequence of orbital nodality.

The revTPSS functional is the best-performing meta-GGA for the 1s state, tightly followed by the rSCAN and r2SCAN functionals. MN15-L shows a better overall performance than LDAs for all states; however, the performance of MN15-L is non-systematic like that of DM21.

Hybrid functionals like BHandH and QTP17 have the best overall performance as they explicitly include some fraction of Hartree–Fock exchange. Moreover, both BHandH and QTP17 are mixtures of Hartree–Fock and LDA exchange, leading to good observed accuracy.

DM21 turns out to be only close to exact for the 1s state OEE from H0 to B4+ (see Fig. 2). For He+ to B4+, DM21 also shows good performance for 2p and 3d states. However, over the whole range of investigated species H0 to Kr35+, DM21 exhibits various trend changes and overall inconsistent performance. Thus, one might improve the next generation of the DM21 functional by including more one-electron cations in the training sets for various elements in the periodic table. This might increase the consistency of promising properties of such kind of machine-learned functionals.

We found PBEsol to be the most accurate GGA functional for the 2p and 3d states. PBEsol is also the most accurate GGA functional overall. These findings are interesting to contrast with that of Lehtola, Jónsson, and Jónsson,47 who showed that PBEsol is one of the few functionals whose accuracy improves when PZ-SIC is applied with complex orbitals. Therefore, the development of novel DFAs with reduced one-electron error could be useful for PZ-SIC calculations, as the reduced one-electron errors [Eq. (2)] would affect the numerics of the PZ correction [Eq. (1)] and might alleviate well-known issues with PZ-SIC and PZFLO-SIC discussed in Sec. I.

Note added in proof After the acceptance of this paper, we became aware of a preprint by Lonsdale and Goerigk [132] that includes discussion on excited states of hydrogenic cations.

See supplementary material for exploratory finite element studies of basis set truncation errors in density-functional calculations. Sorted rankings of the functionals by errors for the 1s, 2p, and 3d states as well as the overall error. Bar plots of the errors for all studied functionals. Comparison of the 1s data to the study of Lonsdale and Goerigk56 with additional basis set incompleteness studies. Tables of the calculated total energies for the 1s, 2p, and 3d states for all studied functionals.

We thank Jens Kortus for their valuable comments on the manuscript. S.S. has been funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) - Project No. 421663657 - KO 1924/9-2. This work was part of the OpenSIC project. We thank ZIH Dresden for computational time and support.

The authors have no conflicts to disclose.

Sebastian Schwalbe: Conceptualization (equal); Data curation (lead); Formal analysis (equal); Investigation (equal); Methodology (equal); Project administration (equal); Software (equal); Supervision (equal); Visualization (lead); Writing – original draft (equal); Writing – review & editing (equal). Kai Trepte: Conceptualization (equal); Data curation (supporting); Formal analysis (equal); Investigation (equal); Methodology (equal); Project administration (equal); Software (equal); Validation (lead); Visualization (supporting); Writing – original draft (equal); Writing – review & editing (equal). Susi Lehtola: Conceptualization (lead); Data curation (equal); Formal analysis (equal); Investigation (equal); Methodology (equal); Project administration (equal); Software (equal); Supervision (lead); Validation (equal); Visualization (supporting); Writing – original draft (equal); Writing – review & editing (lead).

The data that support the findings of this study are available within the article and its supplementary material.

1.
P.
Hohenberg
and
W.
Kohn
, “
Inhomogeneous electron gas
,”
Phys. Rev. B
136
,
864
871
(
1964
).
2.
W.
Kohn
and
L. J.
Sham
, “
Self-consistent equations including exchange and correlation effects
,”
Phys. Rev.
140
,
A1133
A1138
(
1965
).
3.
A. D.
Becke
, “
Perspective: Fifty years of density-functional theory in chemical physics
,”
J. Chem. Phys.
140
,
18A301
(
2014
).
4.
R. O.
Jones
, “
Density functional theory: Its origins, rise to prominence, and future
,”
Rev. Mod. Phys.
87
,
897
923
(
2015
).
5.
N.
Mardirossian
and
M.
Head-Gordon
, “
Thirty years of density functional theory in computational chemistry: An overview and extensive assessment of 200 density functionals
,”
Mol. Phys.
115
,
2315
2372
(
2017
).
6.
L.
Goerigk
and
N.
Mehta
, “
A trip to the density functional theory zoo: Warnings and recommendations for the user
,”
Aust. J. Chem.
72
,
563
(
2019
).
7.
U.
von Barth
, “
Basic density-functional theory—An overview
,”
Phys. Scr.
2004
(
T109
),
9
.
8.
A. J.
Cohen
,
P.
Mori-Sánchez
, and
W.
Yang
, “
Challenges for density functional theory
,”
Chem. Rev.
112
,
289
320
(
2012
).
9.
S.
Lehtola
,
C.
Steigemann
,
M. J. T.
Oliveira
, and
M. A. L.
Marques
, “
Recent developments in LIBXC—A comprehensive library of functionals for density functional theory
,”
SoftwareX
7
,
1
5
(
2018
).
10.
J. P.
Perdew
,
K.
Burke
, and
M.
Ernzerhof
, “
Generalized gradient approximation made simple
,”
Phys. Rev. Lett.
77
,
3865
3868
(
1996
).
11.
J.
Tao
,
J. P.
Perdew
,
V. N.
Staroverov
, and
G. E.
Scuseria
, “
Climbing the density functional ladder: Nonempirical meta-generalized gradient approximation designed for molecules and solids
,”
Phys. Rev. Lett.
91
,
146401
(
2003
).
12.
J. P.
Perdew
,
J.
Tao
,
V. N.
Staroverov
, and
G. E.
Scuseria
, “
Meta-generalized gradient approximation: Explanation of a realistic nonempirical density functional
,”
J. Chem. Phys.
120
,
6898
6911
(
2004
).
13.
J.
Sun
,
A.
Ruzsinszky
, and
J. P.
Perdew
, “
Strongly constrained and appropriately normed semilocal density functional
,”
Phys. Rev. Lett.
115
,
036402
(
2015
).
14.
P. J.
Stephens
,
F. J.
Devlin
,
C. F.
Chabalowski
, and
M. J.
Frisch
, “
Ab initio calculation of vibrational absorption and circular dichroism spectra using density functional force fields
,”
J. Phys. Chem.
98
,
11623
11627
(
1994
).
15.
A. D.
Becke
, “
Density-functional thermochemistry. V. Systematic optimization of exchange-correlation functionals
,”
J. Chem. Phys.
107
,
8554
(
1997
).
16.
F. A.
Hamprecht
,
A. J.
Cohen
,
D. J.
Tozer
, and
N. C.
Handy
, “
Development and assessment of new exchange-correlation functionals
,”
J. Chem. Phys.
109
,
6264
(
1998
).
17.
A. D.
Boese
,
N. L.
Doltsinis
,
N. C.
Handy
, and
M.
Sprik
, “
New generalized gradient approximation functionals
,”
J. Chem. Phys.
112
,
1670
(
2000
).
18.
Y.
Zhao
,
N. E.
Schultz
, and
D. G.
Truhlar
, “
Exchange-correlation functional with broad accuracy for metallic and nonmetallic compounds, kinetics, and noncovalent interactions
,”
J. Chem. Phys.
123
,
161103
(
2005
).
19.
Y.
Zhao
,
N. E.
Schultz
, and
D. G.
Truhlar
, “
Design of density functionals by combining the method of constraint satisfaction with parametrization for thermochemistry, thermochemical kinetics, and noncovalent interactions
,”
J. Chem. Theory Comput.
2
,
364
382
(
2006
).
20.
N.
Mardirossian
and
M.
Head-Gordon
, “
How accurate are the Minnesota density functionals for noncovalent interactions, isomerization energies, thermochemistry, and barrier heights involving molecules composed of main-group elements?
,”
J. Chem. Theory Comput.
12
,
4303
4325
(
2016
).
21.
N.
Mardirossian
and
M.
Head-Gordon
, “
ωB97X-V: A 10-parameter, range-separated hybrid, generalized gradient approximation density functional with nonlocal correlation, designed by a survival-of-the-fittest strategy
,”
Phys. Chem. Chem. Phys.
16
,
9904
9924
(
2014
).
22.
N.
Mardirossian
and
M.
Head-Gordon
, “
Mapping the genome of meta-generalized gradient approximation density functionals: The search for B97M-V
,”
J. Chem. Phys.
142
,
074111
(
2015
).
23.
N.
Mardirossian
and
M.
Head-Gordon
, “
ωB97M-V: A combinatorially optimized, range-separated hybrid, meta-GGA density functional with VV10 nonlocal correlation
,”
J. Chem. Phys.
144
,
214110
(
2016
).
24.
K.
Brown
,
Y.
Maimaiti
,
K.
Trepte
,
T.
Bligaard
, and
J.
Voss
, “
MCML: Combining physical constraints with experimental data for a multi-purpose meta-generalized gradient approximation
,”
J. Comput. Chem.
42
,
2004
2013
(
2021
).
25.
Z. M.
Sparrow
,
B. G.
Ernst
,
T. K.
Quady
, and
R. A.
DiStasio
, “
Uniting nonempirical and empirical density functional approximation strategies using constraint-based regularization
,”
J. Phys. Chem. Lett.
13
,
6896
6904
(
2022
).
26.
K.
Trepte
and
J.
Voss
, “
Data-driven and constrained optimization of semi-local exchange and nonlocal correlation functionals for materials and surface chemistry
,”
J. Comput. Chem.
43
,
1104
1112
(
2022
).
27.
M. G.
Medvedev
,
I. S.
Bushmarinov
,
J.
Sun
,
J. P.
Perdew
, and
K. A.
Lyssenko
, “
Density functional theory is straying from the path toward the exact functional
,”
Science
355
,
49
52
(
2017
).
28.
K. P.
Kepp
, “
Comment on ‘Density functional theory is straying from the path toward the exact functional’
,”
Science
356
,
496
(
2017
).
29.
M. G.
Medvedev
,
I. S.
Bushmarinov
,
J.
Sun
,
J. P.
Perdew
, and
K. A.
Lyssenko
, “
Response to Comment on ‘Density functional theory is straying from the path toward the exact functional’
,”
Science
356
,
496
(
2017
).
30.
K. P.
Kepp
, “
Energy vs density on paths toward more exact density functionals
,”
Phys. Chem. Chem. Phys.
20
,
7538
7548
(
2018
).
31.
N. Q.
Su
,
Z.
Zhu
, and
X.
Xu
, “
Doubly hybrid density functionals that correctly describe both density and energy for atoms
,”
Proc. Natl. Acad. Sci. U. S. A.
115
,
2287
2292
(
2018
).
32.
Y.
Wang
,
X.
Wang
,
D. G.
Truhlar
, and
X.
He
, “
How well can the M06 suite of functionals describe the electron densities of Ne, Ne6+, and Ne8+?
,”
J. Chem. Theory Comput.
13
,
6068
6077
(
2017
).
33.
E.
Sim
,
S.
Song
, and
K.
Burke
, “
Quantifying density errors in DFT
,”
J. Phys. Chem. Lett.
9
,
6385
6392
(
2018
); arXiv:1809.10347.
34.
A. J.
Cohen
,
P.
Mori-Sánchez
, and
W.
Yang
, “
Fractional view of the exchange-correlation functional and derivative discontinuity in density functional theory
,” Psi-k Scientific Highlight of the Month 99,
2010
.
35.
K. R.
Bryenton
,
A. A.
Adeleke
,
S. G.
Dale
, and
E. R.
Johnson
, “
Delocalization error: The greatest outstanding challenge in density-functional theory
,”
Wiley Interdiscip. Rev.: Comput. Mol. Sci.
e1631
(
2022
).
36.
A.
Ruzsinszky
,
J. P.
Perdew
,
G. I.
Csonka
,
O. A.
Vydrov
, and
G. E.
Scuseria
, “
Spurious fractional charge on dissociated atoms: Pervasive and resilient self-interaction error of common density functionals
,”
J. Chem. Phys.
125
,
194112
(
2006
).
37.
B. G.
Johnson
,
C. A.
Gonzales
,
P. M. W.
Gill
, and
J. A.
Pople
, “
A density functional study of the simplest hydrogen abstraction reaction. Effect of self-interaction correction
,”
Chem. Phys. Lett.
221
,
100
108
(
1994
).
38.
M. C.
Kim
,
E.
Sim
, and
K.
Burke
, “
Understanding and reducing errors in density functional calculations
,”
Phys. Rev. Lett.
111
,
073003
(
2013
); arXiv:1212.3054.
39.
E.
Sim
,
S.
Song
,
S.
Vuckovic
, and
K.
Burke
, “
Improving results by improving densities: Density-corrected density functional theory
,”
J. Am. Chem. Soc.
144
,
6625
6639
(
2022
).
40.
J. L.
Bao
,
Y.
Wang
,
X.
He
,
L.
Gagliardi
, and
D. G.
Truhlar
, “
Multiconfiguration pair-density functional theory is free from delocalization error
,”
J. Phys. Chem. Lett.
8
,
5616
5620
(
2017
).
41.
J. L.
Bao
,
L.
Gagliardi
, and
D. G.
Truhlar
, “
Self-interaction error in density functional theory: An appraisal
,”
J. Phys. Chem. Lett.
9
,
2353
2358
(
2018
).
42.
J. P.
Perdew
and
A.
Zunger
, “
Self-interaction correction to density-functional approximations for many-electron systems
,”
Phys. Rev. B
23
,
5048
5079
(
1981
).
43.
M. R.
Pederson
,
R. A.
Heaton
, and
C. C.
Lin
, “
Local-density Hartree–Fock theory of electronic states of molecules with self-interaction correction
,”
J. Chem. Phys.
80
,
1972
(
1984
).
44.
S.
Lehtola
and
H.
Jónsson
, “
Variational, self-consistent implementation of the Perdew–Zunger self-interaction correction with complex optimal orbitals
,”
J. Chem. Theory Comput.
10
,
5324
5337
(
2014
).
45.
J. P.
Perdew
,
A.
Ruzsinszky
,
J.
Sun
, and
M. R.
Pederson
, “
Paradox of self-interaction correction: How can anything so right be so wrong?
,”
Adv. At., Mol., Opt. Phys.
64
,
1
14
(
2015
).
46.
S.
Lehtola
,
M.
Head-Gordon
, and
H.
Jónsson
, “
Complex orbitals, multiple local minima, and symmetry breaking in Perdew–Zunger self-interaction corrected density functional theory calculations
,”
J. Chem. Theory Comput.
12
,
3195
3207
(
2016
).
47.
S.
Lehtola
,
E. Ö.
Jónsson
, and
H.
Jónsson
, “
Effect of complex-valued optimal orbitals on atomization energies with the Perdew–Zunger self-interaction correction to density functional theory
,”
J. Chem. Theory Comput.
12
,
4296
4302
(
2016
).
48.
M. R.
Pederson
,
A.
Ruzsinszky
, and
J. P.
Perdew
, “
Communication: Self-interaction correction with unitary invariance in density functional theory
,”
J. Chem. Phys.
140
,
121103
(
2014
).
49.
K.
Trepte
,
S.
Schwalbe
,
S.
Liebing
,
W. T.
Schulze
,
J.
Kortus
,
H.
Myneni
,
A. V.
Ivanov
, and
S.
Lehtola
, “
Chemical bonding theories as guides for self-interaction corrected solutions: Multiple local minima and symmetry breaking
,”
J. Chem. Phys.
155
,
224109
(
2021
).
50.
X.
Cheng
,
Y.
Zhang
,
E.
Jónsson
,
H.
Jónsson
, and
P. M.
Weber
, “
Charge localization in a diamine cation provides a test of energy functionals and self-interaction correction
,”
Nat. Commun.
7
,
11013
(
2016
).
51.
Y.
Zhang
,
P. M.
Weber
, and
H.
Jónsson
, “
Self-interaction corrected functional calculations of a dipole-bound molecular anion
,”
J. Phys. Chem. Lett.
7
,
2068
2073
(
2016
).
52.
A. V.
Ivanov
,
T. K.
Ghosh
,
E. Ö.
Jónsson
, and
H.
Jónsson
, “
Mn dimer can be described accurately with density functional calculations when self-interaction correction is applied
,”
J. Phys. Chem. Lett.
12
,
4240
4246
(
2021
).
53.
E. Ö.
Jónsson
,
S.
Lehtola
, and
H.
Jónsson
, “
Towards an optimal gradient-dependent energy functional of the PZ-SIC form
,”
Procedia Comput. Sci.
51
,
1858
1864
(
2015
).
54.
J.
Sun
,
J. P.
Perdew
,
Z.
Yang
, and
H.
Peng
, “
Communication: Near-locality of exchange and correlation density functionals for 1- and 2-electron systems
,”
J. Chem. Phys.
144
,
191101
(
2016
); arXiv:1603.04062.
55.
C.
Shahi
,
P.
Bhattarai
,
K.
Wagle
,
B.
Santra
,
S.
Schwalbe
,
T.
Hahn
,
J.
Kortus
,
K. A.
Jackson
,
J. E.
Peralta
,
K.
Trepte
,
S.
Lehtola
,
N. K.
Nepal
,
H.
Myneni
,
B.
Neupane
,
S.
Adhikari
,
A.
Ruzsinszky
,
Y.
Yamamoto
,
T.
Baruah
,
R. R.
Zope
, and
J. P.
Perdew
, “
Stretched or noded orbital densities and self-interaction correction in density functional theory
,”
J. Chem. Phys.
150
,
174102
(
2019
).
56.
D. R.
Lonsdale
and
L.
Goerigk
, “
The one-electron self-interaction error in 74 density functional approximations: A case study on hydrogenic mono- and dinuclear systems
,”
Phys. Chem. Chem. Phys.
22
,
15805
15830
(
2020
).
57.
O.
Gunnarsson
and
B. I.
Lundqvist
, “
Exchange and correlation in atoms, molecules, and solids by the spin-density-functional formalism
,”
Phys. Rev. B
13
,
4274
4298
(
1976
).
58.
J.
Kirkpatrick
,
B.
McMorrow
,
D. H. P.
Turban
,
A. L.
Gaunt
,
J. S.
Spencer
,
A. G. D. G.
Matthews
,
A.
Obika
,
L.
Thiry
,
M.
Fortunato
,
D.
Pfau
,
L. R.
Castellanos
,
S.
Petersen
,
A. W. R.
Nelson
,
P.
Kohli
,
P.
Mori-Sánchez
,
D.
Hassabis
, and
A. J.
Cohen
, “
Pushing the frontiers of density functionals by solving the fractional electron problem
,”
Science
374
,
1385
1389
(
2021
).
59.
S.
Lehtola
and
A. J.
Karttunen
, “
Free and open source software for computational chemistry education
,”
Wiley Interdiscip. Rev.: Comput. Mol. Sci.
12
,
e1610
(
2022
).
60.
Q.
Sun
,
X.
Zhang
,
S.
Banerjee
,
P.
Bao
,
M.
Barbry
,
N. S.
Blunt
,
N. A.
Bogdanov
,
G. H.
Booth
,
J.
Chen
,
Z.-H.
Cui
,
J. J.
Eriksen
,
Y.
Gao
,
S.
Guo
,
J.
Hermann
,
M. R.
Hermes
,
K.
Koh
,
P.
Koval
,
S.
Lehtola
,
Z.
Li
,
J.
Liu
,
N.
Mardirossian
,
J. D.
McClain
,
M.
Motta
,
B.
Mussard
,
H. Q.
Pham
,
A.
Pulkin
,
W.
Purwanto
,
P. J.
Robinson
,
E.
Ronca
,
E. R.
Sayfutyarova
,
M.
Scheurer
,
H. F.
Schurkus
,
J. E. T.
Smith
,
C.
Sun
,
S.-N.
Sun
,
S.
Upadhyay
,
L. K.
Wagner
,
X.
Wang
,
A.
White
,
J. D.
Whitfield
,
M. J.
Williamson
,
S.
Wouters
,
J.
Yang
,
J. M.
Yu
,
T.
Zhu
,
T. C.
Berkelbach
,
S.
Sharma
,
A. Y.
Sokolov
, and
G. K.-L.
Chan
, “
Recent developments in the PySCF program package
,”
J. Chem. Phys.
153
,
024109
(
2020
); arXiv:2002.12531.
61.
S.
Lehtola
, “
Assessment of initial guesses for self-consistent field calculations. superposition of atomic potentials: Simple yet efficient
,”
J. Chem. Theory Comput.
15
,
1593
1604
(
2019
); arXiv:1810.11659.
62.
P.
Pulay
, “
Convergence acceleration of iterative sequences. The case of scf iteration
,”
Chem. Phys. Lett.
73
,
393
398
(
1980
).
63.
P.
Pulay
, “
Improved SCF convergence acceleration
,”
J. Comput. Chem.
3
,
556
560
(
1982
).
64.
V. R.
Saunders
and
I. H.
Hillier
, “
A ‘Level-Shifting’ method for converging closed shell Hartree–Fock wave functions
,”
Int. J. Quantum Chem.
7
,
699
705
(
1973
).
65.
S.
Lehtola
, “
Polarized Gaussian basis sets from one-electron ions
,”
J. Chem. Phys.
152
,
134108
(
2020
); arXiv:2001.04224.
66.
F.
Jensen
, “
Polarization consistent basis sets: Principles
,”
J. Chem. Phys.
115
,
9113
9125
(
2001
).
67.
F.
Jensen
, “
Polarization consistent basis sets. III. The importance of diffuse functions
,”
J. Chem. Phys.
117
,
9234
9240
(
2002
).
68.
B. P.
Pritchard
,
D.
Altarawy
,
B.
Didier
,
T. D.
Gibson
, and
T. L.
Windus
, “
New basis set exchange: An open, up-to-date resource for the molecular sciences community
,”
J. Chem. Inf. Model.
59
,
4814
4820
(
2019
).
69.
J.-D.
Chai
and
M.
Head-Gordon
, “
Long-range corrected hybrid density functionals with damped atom-atom dispersion corrections
,”
Phys. Chem. Chem. Phys.
10
,
6615
6620
(
2008
).
70.
A. D.
Becke
, “
Density-functional exchange-energy approximation with correct asymptotic behavior
,”
Phys. Rev. A
38
,
3098
3100
(
1988
).
71.
C.
Lee
,
W.
Yang
, and
R. G.
Parr
, “
Development of the Colle-Salvetti correlation-energy formula into a functional of the electron density
,”
Phys. Rev. B
37
,
785
789
(
1988
).
72.
B.
Miehlich
,
A.
Savin
,
H.
Stoll
, and
H.
Preuss
, “
Results obtained with the correlation energy density functionals of becke and Lee, Yang and Parr
,”
Chem. Phys. Lett.
157
,
200
206
(
1989
).
73.
A. D.
Becke
, “
A new mixing of Hartree–Fock and local density-functional theories
,”
J. Chem. Phys.
98
,
1372
(
1993
).
74.
L. A.
Constantin
,
E.
Fabiano
, and
F.
Della Sala
, “
Meta-GGA exchange-correlation functional with a balanced treatment of nonlocality
,”
J. Chem. Theory Comput.
9
,
2256
(
2013
).
75.
L. A.
Constantin
,
E.
Fabiano
, and
F.
Della Sala
, “
Semilocal dynamical correlation with increased localization
,”
Phys. Rev. B
86
,
035130
(
2012
).
76.
M.
Renz
,
K.
Theilacker
,
C.
Lambert
, and
M.
Kaupp
, “
A reliable quantum-chemical protocol for the characterization of organic mixed-valence compounds
,”
J. Am. Chem. Soc.
131
,
16292
16302
(
2009
).
77.
M.
Kaupp
,
M.
Renz
,
M.
Parthey
,
M.
Stolte
,
F.
Würthner
, and
C.
Lambert
, “
Computational and spectroscopic studies of organic mixed-valence compounds: Where is the charge?
,”
Phys. Chem. Chem. Phys.
13
,
16973
16986
(
2011
).
78.
T.
Tsuneda
,
T.
Suzumura
, and
K.
Hirao
, “
A new one-parameter progressive Colle–Salvetti-type correlation functional
,”
J. Chem. Phys.
110
,
10664
(
1999
).
79.
T.
Yanai
,
D. P.
Tew
, and
N. C.
Handy
, “
A new hybrid exchange-correlation functional using the Coulomb-attenuating method (CAM-B3LYP)
,”
Chem. Phys. Lett.
393
,
51
(
2004
).
80.
P.
Verma
and
R. J.
Bartlett
, “
Increasing the applicability of density functional theory. IV. Consequences of ionization-potential improved exchange-correlation potentials
,”
J. Chem. Phys.
140
,
18A534
(
2014
).
81.
Y.
Jin
and
R. J.
Bartlett
, “
The QTP family of consistent functionals and potentials in Kohn–Sham density functional theory
,”
J. Chem. Phys.
145
,
034107
(
2016
).
82.
R. L. A.
Haiduke
and
R. J.
Bartlett
, “
Non-empirical exchange-correlation parameterizations based on exact conditions from correlated orbital theory
,”
J. Chem. Phys.
148
,
184106
(
2018
).
83.
P. A. M.
Dirac
, “
Note on exchange phenomena in the Thomas atom
,”
Math. Proc. Cambridge Philos. Soc.
26
,
376
(
1930
).
84.
T.
Chachiyo
, “
Simple and accurate uniform electron gas correlation energy for the full range of densities
,”
J. Chem. Phys.
145
,
021101
(
2016
).
85.
H. S.
Yu
,
W.
Zhang
,
P.
Verma
,
X.
He
, and
D. G.
Truhlar
, “
Nonseparable exchange-correlation functional for molecules, including homogeneous catalysis involving transition metals
,”
Phys. Chem. Chem. Phys.
17
,
12146
12160
(
2015
).
86.
J.
Heyd
,
G. E.
Scuseria
, and
M.
Ernzerhof
, “
Hybrid functionals based on a screened Coulomb potential
,”
J. Chem. Phys.
118
,
8207
(
2003
).
87.
J.
Heyd
,
G. E.
Scuseria
, and
M.
Ernzerhof
, “
Erratum: ‘Hybrid functionals based on a screened Coulomb potential’ [J. Chem. Phys. 118, 8207 (2003)]
,”
J. Chem. Phys.
124
,
219906
(
2006
).
88.
A. V.
Krukau
,
O. A.
Vydrov
,
A. F.
Izmaylov
, and
G. E.
Scuseria
, “
Influence of the exchange screening parameter on the performance of screened hybrid functionals
,”
J. Chem. Phys.
125
,
224106
(
2006
).
89.
J. E.
Moussa
,
P. A.
Schultz
, and
J. R.
Chelikowsky
, “
Analysis of the Heyd–Scuseria–Ernzerhof density functional parameter space
,”
J. Chem. Phys.
136
,
204117
(
2012
).
90.
O. A.
Vydrov
and
T.
Van Voorhis
, “
Nonlocal van der Waals density functional: The simpler the better
,”
J. Chem. Phys.
133
,
244103
(
2010
).
91.
M. A.
Rohrdanz
,
K. M.
Martins
, and
J. M.
Herbert
, “
A long-range-corrected density functional that performs well for both ground-state properties and time-dependent density functional theory excitation energies, including charge-transfer excited states
,”
J. Chem. Phys.
130
,
054112
(
2009
).
92.
Y.
Zhao
and
D. G.
Truhlar
, “
A new local density functional for main-group thermochemistry, transition metal bonding, thermochemical kinetics, and noncovalent interactions
,”
J. Chem. Phys.
125
,
194101
(
2006
).
93.
Y.
Zhao
and
D. G.
Truhlar
, “
The M06 suite of density functionals for main group thermochemistry, thermochemical kinetics, noncovalent interactions, excited states, and transition elements: Two new functionals and systematic testing of four M06-class functionals and 12 other functionals
,”
Theor. Chem. Acc.
120
,
215
(
2008
).
94.
R.
Peverati
and
D. G.
Truhlar
, “
M11-L: A local density functional that provides improved accuracy for electronic structure calculations in chemistry and physics
,”
J. Phys. Chem. Lett.
3
,
117
(
2012
).
95.
R.
Peverati
and
D. G.
Truhlar
, “
An improved and broadly accurate local approximation to the exchange-correlation density functional: The MN12-L functional for electronic structure calculations in chemistry and physics
,”
Phys. Chem. Chem. Phys.
14
,
13171
(
2012
).
96.
H. S.
Yu
,
X.
He
,
S. L.
Li
, and
D. G.
Truhlar
, “
MN15: A Kohn–Sham global-hybrid exchange-correlation density functional with broad accuracy for multi-reference and single-reference systems and noncovalent interactions
,”
Chem. Sci.
7
,
5032
5051
(
2016
).
97.
H. S.
Yu
,
X.
He
, and
D. G.
Truhlar
, “
MN15-L: A new local exchange-correlation functional for Kohn–Sham density functional theory with broad accuracy for atoms, molecules, and solids
,”
J. Chem. Theory Comput.
12
,
1280
1293
(
2016
).
98.
J.
Sun
,
B.
Xiao
, and
A.
Ruzsinszky
, “
Communication: Effect of the orbital-overlap dependence in the meta-generalized gradient approximation
,”
J. Chem. Phys.
137
,
051101
(
2012
).
99.
J. P.
Perdew
,
A.
Ruzsinszky
,
G. I.
Csonka
,
L. A.
Constantin
, and
J.
Sun
, “
Workhorse semilocal density functional for condensed matter physics and quantum chemistry
,”
Phys. Rev. Lett.
103
,
026403
(
2009
).
100.
J. P.
Perdew
,
A.
Ruzsinszky
,
G. I.
Csonka
,
L. A.
Constantin
, and
J.
Sun
, “
Erratum: Workhorse semilocal density functional for condensed matter physics and quantum chemistry [Phys. Rev. Lett. 103, 026403 (2009)]
,”
Phys. Rev. Lett.
106
,
179902
(
2011
).
101.
J.
Sun
,
R.
Haunschild
,
B.
Xiao
,
I. W.
Bulik
,
G. E.
Scuseria
, and
J. P.
Perdew
, “
Semilocal and hybrid meta-generalized gradient approximations based on the understanding of the kinetic-energy-density dependence
,”
J. Chem. Phys.
138
,
044113
(
2013
).
102.
N. C.
Handy
and
A. J.
Cohen
, “
Left-right correlation energy
,”
Mol. Phys.
99
,
403
(
2001
).
103.
J. P.
Perdew
,
K.
Burke
, and
M.
Ernzerhof
, “
Errata: Generalized gradient approximation made simple [Phys. Rev. Lett. 77, 3865 (1996)]
,”
Phys. Rev. Lett.
78
,
1396
(
1997
).
104.
J. P.
Perdew
,
A.
Ruzsinszky
,
G. I.
Csonka
,
O. A.
Vydrov
,
G. E.
Scuseria
,
L. A.
Constantin
,
X.
Zhou
, and
K.
Burke
, “
Restoring the density-gradient expansion for exchange in solids and surfaces
,”
Phys. Rev. Lett.
100
,
136406
(
2008
).
105.
J. P.
Perdew
,
S.
Kurth
,
A.
Zupan
, and
P.
Blaha
, “
Accurate density functional with correct formal properties: A step beyond the generalized gradient approximation
,”
Phys. Rev. Lett.
82
,
2544
(
1999
).
106.
P.
Ziesche
and
H.
Eschrig
, in
Electronic Structure of Solids’ 91: Proceedings of the 75. WE-Heraeus-Seminar and 21st Annual International Symposium on Electronic Structure of Solids Held in Gaussig (Germany), March 11–15, 1991
(
De Gruyter Akademie Forschung
,
1991
), Vol. 17.
107.
J. P.
Perdew
,
J. A.
Chevary
,
S. H.
Vosko
,
K. A.
Jackson
,
M. R.
Pederson
,
D. J.
Singh
, and
C.
Fiolhais
, “
Atoms, molecules, solids, and surfaces: Applications of the generalized gradient approximation for exchange and correlation
,”
Phys. Rev. B
46
,
6671
(
1992
).
108.
J. P.
Perdew
,
J. A.
Chevary
,
S. H.
Vosko
,
K. A.
Jackson
,
M. R.
Pederson
,
D. J.
Singh
, and
C.
Fiolhais
, “
Erratum: Atoms, molecules, solids, and surfaces: Applications of the generalized gradient approximation for exchange and correlation
,”
Phys. Rev. B
48
,
4978
(
1993
).
109.
Y.
Jin
and
R. J.
Bartlett
, “
Accurate computation of X-ray absorption spectra with ionization potential optimized global hybrid functional
,”
J. Chem. Phys.
149
,
064111
(
2018
).
110.
B.
Hammer
,
L. B.
Hansen
, and
J. K.
Nørskov
, “
Improved adsorption energetics within density-functional theory using revised Perdew–Burke–Ernzerhof functionals
,”
Phys. Rev. B
59
,
7413
(
1999
).
111.
J. P.
Perdew
and
Y.
Wang
, “
Accurate and simple analytic representation of the electron-gas correlation energy
,”
Phys. Rev. B
45
,
13244
(
1992
).
112.
F.
Bloch
, “
Bemerkung zur Elektronentheorie des Ferromagnetismus und der elektrischen Leitfähigkeit
,”
Z. Phys.
57
,
545
(
1929
).
113.
S. H.
Vosko
,
L.
Wilk
, and
M.
Nusair
, “
Accurate spin-dependent electron liquid correlation energies for local spin density calculations: A critical analysis
,”
Can. J. Phys.
58
,
1200
(
1980
).
114.
T.
Aschebrock
and
S.
Kümmel
, “
Ultranonlocality and accurate band gaps from a meta-generalized gradient approximation
,”
Phys. Rev. Res.
1
,
033082
(
2019
).
115.
J.
Tao
and
Y.
Mo
, “
Accurate semilocal density functional for condensed-matter physics and quantum chemistry
,”
Phys. Rev. Lett.
117
,
073001
(
2016
).
116.
V. N.
Staroverov
,
G. E.
Scuseria
,
J.
Tao
, and
J. P.
Perdew
, “
Comparative assessment of a new nonempirical density functional: Molecules and hydrogen-bonded complexes
,”
J. Chem. Phys.
119
,
12129
(
2003
).
117.
X.
Xu
and
W. A.
Goddard
, “
The X3LYP extended density functional for accurate descriptions of nonbond interactions, spin states, and thermochemical properties
,”
Proc. Natl. Acad. Sci. U. S. A.
101
,
2673
(
2004
).
118.
J. W.
Furness
,
A. D.
Kaplan
,
J.
Ning
,
J. P.
Perdew
, and
J.
Sun
, “
Accurate and numerically efficient r2SCAN meta-generalized gradient approximation
,”
J. Phys. Chem. Lett.
11
,
8208
8215
(
2020
).
119.
J. W.
Furness
,
A. D.
Kaplan
,
J.
Ning
,
J. P.
Perdew
, and
J.
Sun
, “
Correction to ‘Accurate and numerically efficient r2SCAN meta-generalized gradient approximation’
,”
J. Phys. Chem. Lett.
11
,
9248
(
2020
).
120.
M.
Bursch
,
H.
Neugebauer
,
S.
Ehlert
, and
S.
Grimme
, “
Dispersion corrected r2SCAN based global hybrid functionals: r2SCANh, r2SCAN0, and r2SCAN50
,”
J. Chem. Phys.
156
,
134105
(
2022
).
121.
A. P.
Bartók
and
J. R.
Yates
, “
Regularized SCAN functional
,”
J. Chem. Phys.
150
,
161101
(
2019
).
122.
Y.
Zhang
and
W.
Yang
, “
Comment on ‘Generalized gradient approximation made simple’
,”
Phys. Rev. Lett.
80
,
890
(
1998
).
123.
J. P.
Perdew
and
K.
Schmidt
, “
Jacob’s ladder of density functional approximations for the exchange-correlation energy
,”
AIP Conf. Proc.
577
,
1
20
(
2001
).
124.
A. D.
Kaplan
,
B.
Santra
,
P.
Bhattarai
,
K.
Wagle
,
S. T. U. R.
Chowdhury
,
P.
Bhetwal
,
J.
Yu
,
H.
Tang
,
K.
Burke
,
M.
Levy
, and
J. P.
Perdew
, “
Simple hydrogenic estimates for the exchange and correlation energies of atoms and atomic ions, with implications for density functional theory
,”
J. Chem. Phys.
153
,
074114
(
2020
).
125.
K. A.
Christensen
and
F.
Jensen
, “
The basis set convergence of the density functional energy for H2
,”
Chem. Phys. Lett.
317
,
400
403
(
2000
).
126.
S.
Lehtola
, “
A review on non-relativistic, fully numerical electronic structure calculations on atoms and diatomic molecules
,”
Int. J. Quantum Chem.
119
,
e25968
(
2019
); arXiv:1902.01431.
127.
S.
Lehtola
, “
Fully numerical Hartree–Fock and density functional calculations. I. Atoms
,”
Int. J. Quantum Chem.
119
,
e25945
(
2019
); arXiv:1810.11651.
128.
S.
Lehtola
, “
Fully numerical calculations on atoms with fractional occupations and range-separated exchange functionals
,”
Phys. Rev. A
101
,
012516
(
2020
); arXiv:1908.02528.
129.
T. H.
Dunning
, “
Gaussian basis sets for use in correlated molecular calculations. I. The atoms boron through neon and hydrogen
,”
J. Chem. Phys.
90
,
1007
(
1989
).
130.
R. A.
Kendall
,
T. H.
Dunning
, and
R. J.
Harrison
, “
Electron affinities of the first-row atoms revisited. Systematic basis sets and wave functions
,”
J. Chem. Phys.
96
,
6796
(
1992
).
131.
N.
Mardirossian
and
M.
Head-Gordon
, “
Characterizing and understanding the remarkably slow basis set convergence of several Minnesota density functionals for intermolecular interaction energies
,”
J. Chem. Theory Comput.
9
,
4453
4461
(
2013
).
132.
D. R.
Lonsdale
and
L.
Goerigk
, “
One-electron self-interactionerror and its relationship to geometry and higher orbital occupation
,” (
2022
).

Supplementary Material