WorksPhysics and Astronomy
olLOSC: Unified and efficient density functional approximation to correct delocalization error in molecules and periodic materials
Abstract
Density functional theory (DFT) is the most promising method for calculating quantum properties of molecules and materials at moderate and large scales. However, commonly used density functional approximations (DFAs) have systematic delocalization error, as demonstrated by underestimated band gaps, over-delocalized charges, and energy level misalignment at interfaces, which limits its quantitative prediction. Extensive efforts, such as the approximation to many-body perturbation theory, system-specific tuning of DFA parameters, and correction functionals have been developed to address delocalization error. However, an accurate, efficient, and unified solution to describe total energy, charge density and band structure for both finite systems and materials is still not available. Building on the linear-response localized orbital scaling correction (lrLOSC), we introduce olLOSC: a localized orbital scaling correction with curvature calculated by orbital-free electronic linear response. olLOSC has comparable accuracy to lrLOSC, but is much more computationally efficient. olLOSC corrects delocalization error - especially underestimated gaps, but also the total energy - both in molecules and in materials with small and moderate band gaps, within the same orbital-free approximation. Critically, with a a unified approximation, olLOSC opens the path for robust and efficient DFT applications across molecules, materials, and interfaces.
main.pdf
Significance statement
Key properties of metals in biology and for developing advanced technologies like catalysts and solar energy converters depend on the quantum-mechanical
Introduction
Density functional theory (DFT) is the “workhorse of quantum chemistry and materials science” [1] and probably the method used most often to predict quantum properties of electronic systems. Among its greatest appeals is its reasonable accuracy at calculating ground-state properties, especially the total energy [2], of molecules and materials.
Kohn–Sham (KS) DFT [3] additionally provides single-particle orbital energies, whose meaning has long been a subject of study. Janak [4] showed that orbital energies computed by a density functional approximation (DFA) are equal to the derivative of the ground-state energy with respect to the orbital occupations, as long as the exchange-correlation functional is continuous in the density . However, this does not connect directly to physical observables. Two relationships connecting observables and orbital energies have been established rigorously. First, the ionization potential theorem [5, 6] demonstrates that, for the exact (local) potential, the highest occupied KS orbital (HOMO) energy is the negative of the first ionization energy (valence band edge) . Second, the ground state chemical potential theorem [7, 8] shows that, in a ground state KS calculation with continuous in or a generalized KS Kohn–Sham (GKS) calculation with continuous in the KS density matrix , the energy of the highest occupied (lowest unoccupied) molecular orbital is the chemical potential of electron removal (addition). The PPLB linearity condition [5] proves that computed by the exact functional are derivatives of the total energy with respect to the number of electrons; thus, for an -electron system,
The frontier orbital energies thus correspond—in both KS and GKS calculations—to experimentally observable ionization potentials and electron affinities (in materials, band edge energies and band gaps). The ground state chemical potential established for the first time the physical meaning of the lowest unoccupied molecular orbital (LUMO) and conduction band minimum (CBM) orbital energy in ground-state (G)KS calculations [7]. DFT calculations thus provide estimates of key properties like the fundamental gap, defined as the difference between the ionization potential and the electron affinity ,
as long as is continuous in or . In molecules, the DFT-computed gap is the difference between the LUMO and HOMO energies; in materials, the difference is between the analogous conduction band minimum (CBM) and valence band maximum (VBM) energies. If the exact density functional were known and used, this result would be exact [4, 5, 7]. Since it is not, (1) is only approximate, and traditional approaches to its approximation are quite poor. Common density functional approximations (DFAs) like LDA [3, 9, 10] and PBE [11] systematically underestimate the fundamental gap by as much as 40% [12]. This problem, as well the underestimation of chemical reaction energy barriers [13, 14], unphysically delocalized charge densities [15], and DFAs’ qualitative failure to describe molecular dissociation [16, 17], are now known to stem from delocalization error [15, 18, 19].
The manifestation of delocalization error depends on the size of the system [20, 21]. In small molecules near their equilibrium bond lengths, it arises from the failure of the energy of a DFA, as a function of the number of electrons, to obey the PPLB linearity condition [5, 22, 23]
In the exact functional, is piecewise linear, with discontinuities in the derivative at integer . Because the derivatives of with respect to the Kohn–Sham orbital occupations are the orbital energies [4], the derivative discontinuities give rise to the expression for the fundamental gap (1). is convex instead of piecewise linear, however, in almost all DFAs used in practice. This convexity softens the derivative discontinuity and, ultimately, causes the litany of problems due to delocalization error. (Note that Hartree–Fock theory suffers instead from localization error because its curve is concave [18]. The success of hybrid functionals, which mix a fraction of Hartree–Fock exchange with a DFA counterpart, is partially due to this cancellation of errors.)
Small molecules exhibit a pronounced curvature in , but its prominence decreases for larger systems. In the bulk limit, translational symmetry forces to be piecewise linear, but the derivative discontinuities (hence band gaps) are still underestimated. In fact, the total energy becomes inaccurate in materials [18]. Because of these contrasting behaviors, it is challenging to correct delocalization error in both molecules and materials with the same approximation.
Kohn–Sham DFT’s underestimation of the (band) gap was perhaps the earliest hallmark of delocalization error to be recognized [12, 15], so there are several fairly mature theories that improve gaps and band structure. However, a method that corrects delocalization error in all scenarios must not only correct the band structure; it must also be able to change the total energy of the system in order to describe the dissociation limit correctly and to correct the charge density to describe interfacial systems [24, 25].
Methods to correct delocalization error
Range-separated hybrid functionals can mitigate delocalization error in molecules or materials, but they do so in completely different ways. In molecules, Coulomb-attenuated functionals [26] use Hartree–Fock (exact) exchange at long range; they cancel delocalization error imperfectly, but improve molecular fundamental gap calculations significantly [27, 28]. In bulk systems, however, the asymptotic decay of Hartree–Fock exchange is qualitatively wrong. Long-range correlations decay exponentially in materials with a band gap [29, 30, 31], and exchange is further scaled by the inverse macroscopic dielectric constant [32]. (More generally, it is scaled by the inverse of the microscopic dielectric function, ; this approach is used in the method [33, 34].) Range-separated hybrids for materials therefore use Hartree–Fock exchange only for short-range interactions, with a fixed screening to describe long-range interactions [35, 36, 37]. The correct mixture of DFT and Hartree–Fock exchange also depends on the particular system. The optimally tuned range-separated hybrid functional [38, 39] incorporates both system-dependent exchange and to correct delocalization error. With its recent extension to materials [40], it is effective at restoring the linearity condition, but the optimal tuning depends not only on the system but on its geometry; it is therefore difficult to obtain consistent energetics of (for instance) the reactants and products a chemical reaction, or of an interface and its separate components. Optimally tuned range-separated hybrid functionals are used primarily to correct band gaps, not total energies.
Koopmans-compliant functionals handily correct quasiparticle spectra in molecules [41, 42]. So does its periodic extension, the Koopmans spectral, or Koopmans-compliant Wannier (KCW) functional [43, 44]. KCW is currently implemented non-self-consistently [45], but because it corrects the Hamiltonian it could be applied self-consistently to provide a density correction, following a similar prescription as [46]. The Wannier–Koopmans method, which shares a similar philosophy, is similarly effective at improving band gap prediction [47–51]. However, Koopmans-compliant corrections do not correct the total energy of insulators or typical molecules.
The method [33, 34] improves gaps and band structures significantly, especially when coupled with some degree of self-consistency [52–55]. Furthermore, it offers an implicit correction to the density. However, its description of the total energy is suspect [56] unless computed fully self-consistently, which is generally prohibitive in computational cost. This scales naively with the fourth power of the number of electrons; even after approximations to reduce the scaling [57], the prefactor is characteristically large, so fully self-consistent calculations are rare. Keeping delocalization error corrections at DFT cost, , with reasonable practical time cost, is thus desirable.
Both the Perdew–Zunger self-interaction correction (PZ-SIC) [58] and its modern descendant, the Fermi–Löwdin orbital self-interaction correction (FLO-SIC) correct delocalization error partially, focusing on its single-electron manifestation. Unlike its earlier predecessor, FLOSIC’s energy correction is invariant under unitary transformation of the occupied orbitals and size-consistent [59]. In addition to energy corrections, FLOSIC can be applied self-consistently [60] and yields a corrected density [61]. However, the quality of its correction to the orbital energies is less clear. FLOSIC provides improvement over DFAs for molecular ionization potentials [62, 63] and, when maximally localized Wannier functions [64, 65] are used to construct the Fermi orbitals, modest improvement in band gaps of semiconductors [66]. However, the improvements are somewhat inconsistent, and worsen when the DFA is a gradient-corrected functional.
The localized orbital scaling method (LOSC), introduced by [25], is a delocalization error correction that offers a correction to the total energy and electron density as well as to band gaps. It does this through the construction of localized orbitalets (called dually localized Wannier functions in materials [67]) that encode both spatial and spectral localization in a balanced way [25] [68]. The original implementation of LOSC performs well for molecular systems, but breaks down in bulk materials because it lacks a description of dielectric screening [69]. More recently, we developed linear-response (lr)LOSC, adding dielectric screening from linear-response theory [21] to LOSC. lrLOSC yields semiconductor and insulator band structures [70], molecular ionization potentials and electron affinities [71], and core-level binding energies [72] of comparable accuracy to (and in some cases better than) methods. Nevertheless, the linear-response kernel is expensive to compute, compared with the conventional DFT calculations, preventing lrLOSC from being broadly applicable to complex bulk systems and interfaces of interest.
In this work, we present olLOSC, a computationally efficient approximation to lrLOSC with an orbital-free linear-response curvature. olLOSC is shown as a unified functional approximation to correct delocalization error in both molecules and semiconductors. Looking ahead, we seek an approximation that makes DFT reliably accurate for computing the electronic structure of interfaces, which combine challenging features of both molecules and materials. olLOSC, applicable to both finite and bulk systems—each a limiting case of the more general interface—is an important step toward this goal.
Theory
All methods in the LOSC family correct delocalization error by applying a quadratic-to-linear correction functional to restore (2) approximately [21, 25] [73, 74]. The most advanced form is the LOSC correction
where the matrix of local occupations is the one-electron density matrix of the noninteracting reference system in a basis of localized orbitalets [25],
The curvature measures the magnitude of the correction for each pair of orbitalets. is the Kronecker delta function, equal to 1 when and to 0 otherwise. is quadratic in , which addresses the major contribution to delocalization error [75].1
The form of shares similarities to those of other well-known post-DFT methods, such as the rotationally invariant formulation [76] of DFT+ [77] and the Koopmans-compliant functionals [41] [78]. The curvature is the second derivative of the total energy with respect to the orbital occupation numbers [21] [79], expressed in the basis of localized orbitalets. It bears strong similarities to linear-response DFT+ [80] and to the Koopmans-compliant Wannier functional [43, 44] [45]. Neither of the latter energy corrections, however, can be applied to molecules and materials in all situations because their respective localized orbitals have fixed occupations, either 1 or 0. Studies on homolytic dissociation demonstrate that the total energy from DFAs is typically accurate near equilibrium, but yields a qualitatively incorrect dissociation limit [15]. DFT+, based on local (often atomic d) orbital occupations, may provide a large correction to the energy even at molecular equilibrium geometries. On the other hand, the Koopmans-compliant energy correction is based on the occupations of Kohn–Sham orbitals or maximally localized Wannier functions, so it can offer no total energy correction in insulating systems, even stretched molecules. In particular, they fail to predict the correct dissociation of molecular ions such as H. A unified correction to delocalization error therefore requires dynamic localization, which is LOSC’s first key feature.
Localization and orbitalets
LOSC’s localized orbitals mix the occupied and virtual manifolds to balance localization in the spatial and energy domains. The resulting orbitals, known as orbitalets for finite systems [25] [68] and dually localized Wannier functions (DLWFs) for periodic systems [67] are obtained from the canonical (Kohn–Sham) orbitals by unitary transformation,
with chosen to minimize a cost function . Initially defined as a penalty based on the Kohn–Sham eigenvalues [25], was reformulated in [68] to take the form
LOSC for materials is implemented in periodic boundary conditions and supports nontrivial sampling of the Brillouin zone; for simplicity, we omit the associated index in this text. We additionally require translational symmetry of orbitals in materials, making them (generalized) Wannier functions [64, 65, 67].
is thus a convex sum of the spatial variance of the orbitalets and their variance in energy [68, 81]. The constant ensures that is dimensionally consistent.2
The key difference between orbitalets (DLWFs) and all previous localization approaches is that they are not limited to mixing orbitals with the same occupation; that is, they allow mixing between occupied (valence) and virtual (conduction) orbitals. Allowing this mixing means that the diagonal occupations are not constrained to integral values, even in insulators. It is these noninteger local occupations that allow LOSC to correct the total energy, necessary for correcting delocalization error in molecular dissociation.
Molecular orbitalets are dynamic: different molecular environments yield qualitatively different orbitalets when is minimized. This is illustrated dramatically by stretched diatomic molecules. Near equilibrium, the relatively large gap between Kohn–Sham eigenvalues forces the orbitalets to approximate the canonical orbitals closely; the local and canonical occupations are essentially the same—integers—and LOSC yields almost no correction to the total energy. If the molecule is stretched, however, the canonical orbitals become increasingly delocalized, and the energy gap between them narrows. The coupled change in spatial and spectral characteristics drives a transition that localizes the orbitalets on their respective atoms. Their occupations become fractional, and LOSC provides a substantial correction to the total energy that improves the description of dissociation greatly [25].
Linear-response curvature
The curvature measures the deviation from linearity of the DFA energy with respect to orbital occupation numbers. Yang and coworkers derived this quantity to second order with respect to the Kohn–Sham (canonical) occupations [79], finding that
Here, is the Hartree–exchange-correlation (Hxc) kernel, and is the (static) linear response function of the density to an external perturbing potential. For details of multiple derivations of , see the Supporting Information of [21].
However, as mentioned above, local orbitals with fractional occupations are required to correct the total energy, necessary to correct delocalization error size-consistently. The linear-response (lr)LOSC curvature ansatz is (7) expressed in the orbitalet basis; for collinear spins, this yields [70, 72]
The linear response function can be written in terms of the noninteracting linear response function with the Dyson equation
where repeated spins (and their corresponding spatial variables) are integrated over. In Kohn–Sham calculations, takes the well-known form
where is the effective potential, which includes both Hxc and external terms. In the second equality, indexes occupied and virtual Kohn–Sham orbitals.
Eq. (10) suggests a self-consistent method for calculating , but since depends on two spatial coordinates and includes a sum over many unoccupied orbitals, such a method promises to be computationally expensive. It turns out that either self-consistency or the sum over unoccupied states can be avoided. We can rewrite the Dyson equation for (omitting spin indices) as
If can be inverted numerically—using, for example, a discretization enabled by the resolution-of-the-identity (RI) method [82]—then can be computed noniteratively. This is the method used in lrLOSC for molecules [71, 72]. On the other hand, the Sternheimer equation [83] can be solved in the manner of density functional perturbation theory, bypassing the sum over virtual orbitals; this method is used in lrLOSC for materials [70].3
Computing with linear response yields a very accurate correction to delocalization error for molecules, semiconductors, and insulators. lrLOSC predicts band gaps ranging from less than 0.5 eV to more than 15 eV with accuracy comparable to or better than self-consistent GW [70]; the fundamental gaps and core-level binding energies of molecules are of similar accuracy [72]. However, even with the speedup afforded by the Sternheimer equation, calculating curvature in materials with linear response remains computationally costly, limiting the application of lrLOSC to crystals with relatively small unit cells compared to systems of practical interest for commonly used DFT approximations.
Orbital-free linear-response curvature
There is another way to think about and that enables a faster calculation with only a modest loss in accuracy: the orbital-free approach [84]. The total energy obeys the stationarity condition
where is the Hohenberg–Kohn universal functional, is the external potential, and is the spin- chemical potential. Taking a second variation yields
and rearranging gives the Euler–Lagrange equation
for the perturbing potential and its linear response [84]. The Lagrange multiplier ensures the conservation of charge (equivalently, electron number). We quickly obtain an expression for the linear-response function:
Finally, in the Kohn–Sham auxiliary system, there is no Hartree–exchange-correlation energy, so is exactly the noninteracting kinetic energy . Thus
If we had the exact orbital-free kinetic energy functional , (9) and (16) would lead to a identical to the coupled-perturbed Kohn–Sham equations. This relationship suggests a computationally simplification; we approximate by the kernel of an orbital-free kinetic energy functional. In this work, we choose the Thomas–Fermi functional [85, 86] with the von Weizsäcker correction [87], obtaining
where is the Thomas–Fermi kernel, the von Weizsäcker kernel, and controls the amount of von Weizsäcker correction. (The analytic form of can be found in the Supporting Information.) Like its Kohn–Sham counterpart, is diagonal in spin, with . The Thomas–Fermi kernel is local in space, while includes derivatives of the Dirac delta distribution; however, the action of on can still be evaluated locally. We thus obtain an orbital-free approximation
to . Replacing by 4 in (9) yields the orbital-free approximation to the interacting response function; solving for directly or iteratively, we obtain the orbital-free linear-response (ol)LOSC curvature .
The lrLOSC and olLOSC curvature expressions contain two components: a bare response to , indicated by the first summand in (8), and a response screened by or from the second summand. It is this screening, or orbital relaxation, that olLOSC approximates with an orbital-free kinetic energy kernel. Because screening is a collective, many-electron effect, using in olLOSC should capture the effect well; choosing a better would likely improve accuracy.
The partial random phase approximation
We also make a partial random phase approximation (RPA) in olLOSC, replacing the Hxc kernel by the Hartree kernel in (9). There are two reasons for doing so. First, including the exchange-correlation kernel in induces major numerical instabilities when computing in materials, while can be obtained by inverting a positive semidefinite matrix. Second, numerical tests on molecules actually demonstrate better fundamental gaps from than from . See the Supporting Information for details. This approximation is akin to the screening described in the approximation [33, 34]: the operator accounts only for Coulombic screening, while exchange-correlation effects are contained in the vertex function (and therefore usually neglected). Note that exchange-correlation interactions are excluded only from , not entirely removed from the olLOSC curvature; is still present in both terms of (8).
Hamiltonian and spectrum
We derive the LOSC correction to the DFA Hamiltonian from (3). At each spin component [25] [46],
Diagonalizing yields corrected canonical orbitals and orbital energies . These eigenvalues are interpreted as quasiparticle energies that produce the LOSC-corrected band structure, including the fundamental gap [7] [88, 89, 90].
Results and Discussion
There are two parameters in olLOSC: the balance between spatial and energy localization and the fraction of von Weizsäcker kinetic energy. We found that and gave the best results when comparing between molecules and materials. Increasing the fraction of von Weizsäcker kinetic energy tends to increase the fundamental gap slightly (decreases the screened response ), while increasing the spatial delocalization has the opposite effect. We note that most previous implementations of the LOSC method use , but in some cases this decreased spatial localization leads the orbitalets of materials to become trapped in local minima (see the Supporting Information for details). All results in this section use . Underlying density functional calculations are performed with the PBE functional [11].
We tested the molecular implementation of olLOSC on the fundamental gaps of small and large molecules and on the ionization potentials of a set of polymers. The small molecules are a 17-member subset of the G2/97 test set [91] such that eV. The reference values for small molecules are fundamental gaps calculated by CCSD(T) in the Supporting Information of [68]. Large molecules are organic acceptors with reference CCSD(T) gaps from [92]. The polymers we study are polyacetylene, H(CH)H, where ; reference ionization potentials are computed with RASPT2 in [93].
The bulk implementation was tested on a set of thirteen semiconductors and large-gap insulators, with lattice constants given by the experimental values found in [37]. The reference values are electronic gaps—experimental gaps adjusted for lattice-coupling effects via zero-point renormalization [94–96];5 they range from less than 1 to more than 21 eV.
As is clear from Fig. 1a, olLOSC provides substantial improvement in molecular ionization potentials and fundamental gaps compared with PBE. With the chosen parameters, we see little systematic error and a greatly improved absolute error. For example, the ionization potential of polyacetylene oligomers (triangles)

Figure 1. DFA (blue) and olLOSC (orange) fundamental gaps (ionization potentials for polyacetylenes), in (a) molecules and polymers, (b) materials. Reference values are from (a) CCSD(T) and RASPT2, (b) experiments, corrected for zero-point renormalization. The orbitalet space/energy mixing parameter , and the kinetic energy functional is TF + 75% vW ().
| MSE (eV) | MAE (eV) | MARE (%) | |
| Molecules | |||
| Small | 0.528 | 0.562 | 5.48 |
| Large | −0.251 | 0.352 | 4.18 |
| All | 0.080 | 0.441 | 4.73 |
| H(CH)H | Polymers (IP only) | ||
| (1 10) | 0.064 | 0.125 | 1.52 |
| Materials | |||
| Gap eV | −0.118 | 0.160 | 5.77 |
| Gap > 8 eV | −2.742 | 2.742 | 20.6 |
| All | −1.329 | 1.352 | 12.6 |
Table 1. Relative errors from olLOSC (, ) of gaps (molecules, materials) and ionization potentials (polymers). Materials’ experimental gaps are renormalized for zero-point energy. Mean signed errors (MSE) and mean absolute errors (MAE) are in eV; mean absolute relative errors (MARE) are percentages.
are both more accurate and better follow the true trendline as their length is increased. Both small and large molecules have far better band gaps. We note a slight overestimation in the gap of molecules with especially large band gaps, in contrast to the situation for bulk insulators (see below).
The band gaps of semiconductors and moderately gapped insulators, up to about 8 eV, are similarly improved (Fig. 1b). For large-gapped insulators (in this dataset, LiCl, NaF, LiF, Ar, and Ne), band gaps are systematically underestimated, although the olLOSC predictions are still better than those of PBE.
The behavior of olLOSC in large-gapped systems differs between molecules and materials. In molecules, we see a small overestimation of the gap; in materials, the opposite effect is observed. Thus, we cannot tune the olLOSC parameters to eliminate the inaccuracy; at best, we can achieve a compromise. We attribute the difference to the limited accuracy of the relatively simple Thomas–Fermi–von Weizsäcker kinetic energy functional. Investigating more sophisticated functionals, such as those discussed in [97], is a promising avenue for future work. However, the TFvW functional is sufficiently accurate in the regime of interest for interfacial systems: inorganic and organic molecules, especially those with modest gaps, and semiconductors. We also expect that olLOSC with will perform adequately when extended to metals, which must account for perturbations to the Fermi energy [98, section II.C.4].
Conclusion
olLOSC corrects delocalization error consistently across system sizes, from small molecules to bulk semiconductors and insulators (with small and moderate band gaps). Besides this, the LOSC family has three crucial ingredients, each necessary for accurately calculating interfaces with DFT: correcting the band structure or quasiparticle energies [69, 70, 72]; modifying the total energy size-consistently [25] [68]; and correcting the charge density [46]. It is equally important to offer corrections to the total energy to describe chemical reactions occurring at interfaces. More fundamentally, approximate density functional calculations with delocalization error are not size-consistent [25], an error that could be exacerbated in interfaces. Finally, changes to the charge density are necessary to capture the charge transfer that can occur across an interface [99–103]. To the best of our knowledge, LOSC is the only method that can satisfy all three requirements in a unified approximation. In future work, we will explore the impact of other kinetic energy functionals [97] and add the long-wavelength correction to the Fermi energy in metals. Finally, we will implement the procedure of [46] in olLOSC to obtain a self-consistently optimized density. We expect that olLOSC will develop into a robust and efficient DFT approach for modeling interfaces between molecules and surfaces.
Materials and methods
Curvature implementation: Molecules
In molecular systems, we discretize and in a low-rank function space inspired by the resolution-of-the-identity (RI) approximation [82, 104–106]; in the resulting auxiliary basis, we can invert them directly. We call our auxiliary, atom-centered basis functions . In the basis, the matrix elements of the orbital-free kinetic energy kernel are
The Hartree kernel is similarly
these matrix elements are computed readily with the RI-V machinery of [82]. Applying this discretization and the partial RPA to solve (11), we obtain the orbital-free linear-response function
As mentioned before, we must impose charge conservation in . However, given , we can conserve charge without computing the chemical potential perturbation directly. The total charge of is
then, following [84], the charge-conserving linear-response function has matrix elements
The olLOSC curvature is thus computed as
where is a Hxc kernel matrix element between the orbitalet and auxiliary bases.
Curvature implementation: Materials
Following [44] [80], we decompose
where the potential and screened density response are given by
We compute and directly. But where is computed in KCW [44] and lrLOSC via perturbations to the occupied Kohn–Sham orbitals, we obtain it more directly in olLOSC. We identify the perturbation of (14) with , yielding the linear system
(Without the partial RPA, we would replace by .) Choosing also fixes the perturbation to the external potential, so that .
The matrix on the left-hand side of (28) is nonlocal in either real or reciprocal space, so it is too large even to store directly. We can, however, compute its action on a vector and use an iterative solver to obtain and simultaneously. Under the partial RPA, the matrix is positive semidefinite, so we solve it with the modified conjugate gradient algorithm of [107]. Instead of directly computing , which changes each iteration, we impose charge conservation as a constraint via a projected preconditioner [107]. Without the partial RPA, a linear solver for indefinite matrices such as MINRES-QLP [108–110] would be required.
Computational details
olLOSC for molecules is implemented in the in-house code , which uses Gaussian-type orbitals. Small and large molecules are computed with correlation-consistent triple-zeta Dunning basis sets augmented with diffuse functions (aug-cc-pVTZ) [111, 112]; the polymers use the same basis sets without augmentation (cc-pVTZ). We use (aug)-cc-pVTZ-RIFIT as the auxiliary basis for .
olLOSC for materials is implemented as a module in a locally maintained fork [113] of the open-source Quantum ESPRESSO package [114, 115], version 7.2. The bulk orbitalets are computed with a fork [116] of wannier90 [117–119] version 3.1.0. Quantum ESPRESSO’s PWscf code is used for the underlying DFT calculations, which use the PBE functional [11] and optimized Vanderbilt norm-conserving pseudopotentials with scalar relativistic corrections [120] downloaded from the ABINIT PseudoDojo (http://www.pseudo-dojo.org/). We use a kinetic energy cutoff of 75 Ry; unless otherwise specified, we sample the Brillouin zone with a Monkhorst–Pack (uniform) grid [121] centered at the origin of reciprocal space. The macroscopic dielectric constant , needed for the Gygi–Baldereschi correction [122] to the Coulomb singularity, is computed by density functional perturbation theory [98] in the Phonon module of Quantum ESPRESSO, with the same parameters. The eigenvalues of the olLOSC Hamiltonian, used for corrected band structures, are computed along the paths specified by [123] for the systems’ respective lattices.
Computational complexity
Computing the curvature—in particular, the screened response—is the most computationally demanding part of a LOSC calculation. Thus, we restrict our complexity analysis to this bottleneck step.
In materials, the monochromatic decomposition means that the density response , which encodes the screening, must be computed times, where is the number of DLWFs and is the number of k-points sampled in the Brillouin zone. Each calculation of requires a matrix inversion, computed by the projected preconditioned conjugate gradient [107]; each iteration scales linearly in the number of plane waves (the length of ) [110]. Each iteration also requires several conversions between real and reciprocal space (because the kinetic energy kernel is semilocal in r former, while the Coulomb kernel is local in G) via the fast Fourier transform, which scales as . Thus, olLOSC for materials scales as . This saves a factor of relative to lrLOSC [70]; even with the Sternheimer equation bypassing direct computation of , constructing in lrLOSC requires coupled pairs of k-points.
In molecules, the interacting linear response function is computed directly; the bottleneck step is inverting and . Within the orbital-free-approximation, all involved matrices are evaluated in a basis of auxiliary functions, where is proportional to the number of orbitalets . We invert with an LU decomposition, so olLOSC for molecules scales as . In addition to the evaluation of the linear response function, the numerical integration of the exchange-correlation kernel can be a potential time-determining step because of its relatively large pre-factor. In olLOSC, this integral is constructed as shown in Equation 25; it scales as , where is the number of grid points. lrLOSC, by comparison, requires the inversion of a rank-4 tensor instead of a rank-2 tensor because (unlike ) depends on the difference between occupied and virtual orbital energies.
Data Availability
The Python (molecules) and R (materials) analysis and figure generation scripts, as well as the underlying data, are included in the Supporting Information. olLOSC for materials is implemented as a fork of Quantum ESPRESSO and is available at [113].
Acknowledgements
We are grateful to Yang Shen for discussion about the positive semidefiniteness of under the partial RPA. We acknowledge financial support from the National Science Foundation (Grant No. CHE-2154831) and the National Institutes of Health (Grant No. R35-GM158181).
Supporting Information for olLOSC: Unified and efficient density functional approximation to correct delocalization error in molecules and periodic materials
Yichen Fan; Jacob Z. Williams; Weitao Yang
supp.pdf
Additional notation for materials
Computing LOSC in periodic boundary conditions requires additional considerations (and more notation). We sample the complex Kohn–Sham orbitals at different points in the irreducible Brillouin zone. This yields an additional index to the canonical orbitals. They also obey Bloch’s theorem [1],
where shares the translational symmetry of the underlying lattice.
The bulk orbitalets must also obey lattice’s translational symmetry. This is shown mathematically by localizing with a different unitary matrix at each -point; the orbitalets are given by
the orbitals and do not mix at different . To obtain a more natural representation in real space, we transform to the (generalized) Wannier representation [2], yielding dually localized Wannier functions (DLWFs) [3]
By the same token, maximally localized Wannier functions [4] are not an exact analogue of the molecular Foster–Boys orbitals [5]; loosening the translational invariance constraint could, in principle, yield localized orbitals with smaller spatial variance. (Of course, the periodic structure of Wannier functions is a major part of their usefulness and appeal.) Note that, in molecules and materials sampled only at , and . This renders the Fourier transformation of (3) trivial, and .
The monochromatic decomposition
It is well known [6, 7] that linear operators that are periodic on a supercell consisting of uniformly sampled primitive unit cells—or, equivalently, on -points sampled in the Brillouin zone—can be decomposed into monochromatic components, each periodic on the primitive cell:
(For the weights of each to be equal, the Brillouin zone must be sampled uniformly [8], and the origin of reciprocal space must be one of the k-points.)
This decomposition applies to one-index quantities such as Bloch orbitals (including their perturbations, useful for density functional perturbation theory); indeed, it is nothing more than a statement of Bloch’s theorem [1]. It also applies to Wannier functions, with
the factor of is a normalization convention.
More recently, Colonna et al. found in [9] that the density of a generalized Wannier function (such as a LOSC DLWF) also has this property:
where
Observe also that
Thus, we only compute quantities for the home unit cell explicitly; for the rest, we require only the phase factors . Because the LOSC curvature operator is linear in the orbital et densities,
we can apply the monochromatic decomposition wholesale, obtaining
The monochromatic decomposition provides a substantial improvement to computational scaling. Naïvely, quantities like and are periodic on the supercell real- and reciprocal-space grids, which have elements (where is the size of the primitive cell grid). Matrix operations on this space scale as . The monochromatic decomposition allows us to work with quantities periodic on the primitive cell, with matrix operations scaling as . Each supercell quantity decomposes into primitive-cell quantities, which means that computing each curvature element scales as overall (up to logarithmic factors due to the fast Fourier transform), saving a factor of . In lrLOSC, the density functional perturbation theory approach [6] uses the Sternheimer equation [10] to compute via the first-order variations of the Kohn–Sham orbitals ; the reciprocal-space points and are paired, so computing in lrLOSC requires effort, saving only one factor of . For more details on the monochromatic decomposition, see [9] and the Supplemental Material of [11].
Implementation of the olLOSC curvature
The Thomas–Fermi–von Weizsäcker kernel
The Thomas–Fermi kinetic energy is
where is the Fermi constant. The variation in with respect to is
and the kernel is
the factor of is omitted for a spinless calculation [12]. The expression for the von Weizsäcker kernel is known, e.g. Eq. (A10) of [13] and Eq. (97) of [14], although the derivation is very long. In Rydberg units, (, but the electron mass ), it is
Together, we have
where is the amount of von Weizsäcker correction. Note that is semilocal; its action on at depends on , , and , but still only requires them to be evaluated at .
olLOSC in molecules
Molecules have few enough basis functions that we can almost invert in (9) directly. The last step is provided by the resolution-of-the-identity (RI) approximation [15, 16] with auxiliary basis functions. Within the RI approximation, the four-center Coulomb integral
can be evaluated in terms of the products of three-center and two-center integrals:
Here, we define
and
here,
label the auxiliary basis functions .
We also use the same set of auxiliary basis functions to discretize the Thomas–Fermi–von Weizsäcker (TFvW) kinetic energy kernel introduced in Equation 14, following the discussion in York and Yang [17]. In our molecular implementation, we evaluate the TFvW kernel in the auxiliary basis and store the resulting coefficients, which are
In the random phase approximation, the inverse of the many-body static linear response function is
We can evaluate the matrix elements of as the sum of the kinetic kernel (21) and Hartree kernel (18). We also implement a beyond-RPA olLOSC scheme, which adds to the right-hand side of (22). For results on this, see 4.3.
olLOSC in materials
To compute the screened density response in materials, we solve the coupled linear equations
Because the Thomas–Fermi–von Weizsäcker kernel is diagonal in spin, this reduces to
Formally, we solve (24) by inverting the left-hand matrix . However, it is nonlocal on either the real-space or reciprocal-space grid, and the number of grid points (plane waves) can exceed . We therefore use a Krylov subspace algorithm to invert it iteratively. Under the partial RPA, is Hermitian and positive semidefinite; as we show in 2.4, for any vector , so we solve for with the conjugate gradient algorithm.
Instead of having to recalculate the charge conservation terms every iteration, we account for them implicitly by constraining the optimization. Including on the right side of (23) is equivalent to requiring
We implement the projected preconditioning method of [18], using a preconditioner that preserves the constraint:
where
is the Thomas–Fermi–von Weizsäcker kernel for the homogeneous electron gas. is the Fermi wavevector given the average density of spin in one unit cell,
is the fraction of von Weizsäcker exchange in ; is the mass of the electron, unity in atomic units but 1/2 in Rydberg units, favored by Quantum ESPRESSO; and is the reduced wavevector magnitude. Note that (and are local in both spin and reciprocal space, with
is positive semidefinite under the partial RPA
We decompose into three components:
By linearity, if , , and are each positive semidefinite, then
and so is .
The Thomas–Fermi kernel
This is the simplest to show. The Thomas–Fermi kernel is
where the constant . (The number of spin components is , here ; and in Rydberg units, where but the electron mass , must be multiplied by 2.)
Given any nonzero two-spin vector ,
since .
The von Weizsäcker kernel
Like the Thomas–Fermi kernel, the vW kernel is diagonal in spin, so . It suffices to consider ; exactly the same argument will apply to the spin- term.
Observe that
where the reduced gradient . Without loss of generality, we ignore the constant .
The derivatives of with respect to and are [19]
Omitting the dependence on ,
Multiplying by powers of to obtain a common denominator, we have
since , (as long as the denominator does not diverge). Replacing the spin index by means that , whence .
The Hartree kernel
We work in reciprocal space, where is local, and note that positive semidefiniteness is invariant under the choice of basis. By the shift theorem of Fourier transforms,
(The divergence at can be handled in various ways; it’s enough to know that .)
The expectation value of any two-spin vector is
We simplify the parts depending on :
so that the total expectation value is the Hartree self-energy of the vector . Since the Hartree kernel is nonnegative (manifest in reciprocal space),
with equality only if .
The LOSC Hamiltonian
Using the chain rule from the generalized complex (or Wirtinger) derivative [20], the LOSC Hamiltonian is given by
with
We can split the correction by spins:
Recall that . We are also assuming that is fixed; this is the frozen-orbitalet approximation for SCF calculation. The orbitalets can be relaxed in the so-called macro-SCF loop, although this has only minor effects [21].
An elementary result from the calculus of variations gives us
similarly, by the Hermiticity of ,
Next, we observe that
and similarly
Thus
To simplify this, we separate diagonal and off-diagonal terms, with
Since the diagonal elements of and are real, the first summand simplifies to
and the second to
(In the above, we used the fact that , with . Observe that the coefficients for each outer product of DLWFs come in complex conjugate pairs.)
Combining, we have
In finite molecules, the Hamiltonian is
with as the localized molecular orbitals.
Choice of parameters
There are two parameters that can be tuned in olLOSC: the space/energy mixing in the orbitalets, and the fraction of von Weizsäcker kinetic energy added to the Thomas–Fermi kernel.
In materials
The results of our parameter sweep are shown in Fig. 1. Since olLOSC is most accurate for systems with smaller gaps ( eV), we chose parameters based on their performance in the left column of the figure. It was clear that some nonzero fraction of von Weizsäcker kinetic energy is required: for all values of the DLWF localization parameter , there was a large negative mean signed error (c, g), indicating systematic underestimation of the band gap.
If , the value used in previous works on LOSC [21, 22, 23, 24], then 100% von Weizsäcker correction is preferred (). However, yields improved performance in some metrics: , has almost zero mean signed percent error (c), while , has minimal mean signed error (eV; g).
We chose for this work to reduce the appearance of local minima in the DLWF minimization (see below), and to minimize the systematic error.
Convergence in Brillouin zone sampling
We investigate the convergence with respect to the k-mesh in Figure 2. (Recall that olLOSC requires a uniform k-sampling centered at for its monochromatic decomposition.) In general, the change in olLOSC gap is fairly small as the k-sampling is increased from to . Noticeable differences appear in Si, C, BP, and SiC because their conduction band minimum is not at a high-symmetry k-point; thus, the olLOSC gap change is matched by a similar one in the PBE calculation.

Figure 2. (a) , (all systems). (b) , (gap eV). (c) , (all systems). (d) , (gap eV). Figure 2: olLOSC convergence with respect to k-points for materials at the two best parameter choices. Left column: All systems. Right column: Small-gapped systems ( eV). Note that NaF and Ne were only computed with a k-mesh.
Large-gapped systems tend to have a slight reduction in the band gap as the k-sampling is increased separate from the DFA. This behavior generally coincides with differences in the ‘degeneracy pattern’ of the DLWFs, as noted in [3]; their convergence is often more difficult in larger-gapped systems. As olLOSC is most applicable to materials with smaller gaps, we did not investigate further. The large deviation in AlP when is also due to a qualitative difference in the DLWFs (see § 16.2 below); nevertheless, it is partially to avoid this possibility that we choose in the main text.
In molecules
The results of our parameter sweep in molecular systems are shown in Figure 3 and Figure 4; is one of 50%, 75%, and 100%, and . Size-consistency is of particular importance in molecular calculations. Small molecules are generally spatially localized but delocalized in energy, have the opposite problem. We therefore partitioned the molecular test based on system size. The small-molecule subset contains the molecules with no more than five atoms; the large-molecule subset contains the rest.

Figure 3. Relative errors (%) by LOSC and the vW in the molecules tested. (a, d): Small molecules. (b, e): Large molecules. (c, f): All molecular systems.

Figure 4. Unscaled errors (eV) by LOSC and the vW in the molecules tested. (a, d): Small molecules. (b, e): Large molecules. (c, f): All molecular systems.
From the standpoint of mean signed error, the molecular datasets allow greater flexibility with respect to . For all three choices of , the fundamental gaps predicted by olLOSC with screening exhibit a mean absolute error of roughly 0.4 eV. Calculations on materials favor larger values, as can be seen in Table 7 (, ) and 10 (, ). Data for the rest of the parameter space can be found in the SI Dataset [25].
| System | (Eh) | (mEh) | HOMO (eV) PBE | HOMO (eV) olLOSC | LUMO (eV) PBE | LUMO (eV) olLOSC | Gap (eV) PBE | Gap (eV) olLOSC | Gap (eV) Ref. | olLOSC gap error eV | olLOSC gap error % |
| Acridine | -555.067 | -0.312 | -5.426 | -7.966 | -2.976 | -0.873 | 2.449 | 7.093 | 7.394 | -0.301 | -4.069 |
| Anthracene | -539.021 | 1.462 | -4.979 | -7.422 | -2.728 | -0.683 | 2.251 | 6.739 | 7.214 | -0.475 | -6.579 |
| Azulene | -385.478 | 8.534 | -4.970 | -7.499 | -2.881 | -0.531 | 2.089 | 6.968 | 7.003 | -0.035 | -0.493 |
| Benzonitrile | -324.201 | -0.087 | -6.852 | -9.924 | -2.521 | 0.069 | 4.331 | 9.992 | 10.168 | -0.176 | -1.728 |
| Benzoquinone | -381.160 | 1.487 | -6.348 | -10.259 | -4.566 | -1.669 | 1.781 | 8.590 | 8.714 | -0.124 | -1.428 |
| Dichlone | -1453.559 | -0.958 | -6.580 | -9.899 | -4.385 | -2.069 | 2.195 | 7.829 | 8.066 | -0.237 | -2.934 |
| Dinitrobenzonitrile | -733.009 | -0.448 | -7.725 | -11.315 | -4.583 | -2.006 | 3.142 | 9.309 | 9.408 | -0.099 | -1.053 |
| Fumaronitrile | -262.854 | -0.312 | -7.851 | -11.310 | -4.094 | -1.183 | 3.757 | 10.128 | 10.507 | -0.379 | -3.611 |
| Maleicanhydride | -379.039 | -0.029 | -7.129 | -11.342 | -4.199 | -1.176 | 2.930 | 10.167 | 10.308 | -0.141 | -1.372 |
| NDCA | -686.067 | 3.383 | -6.468 | -9.023 | -3.658 | -1.394 | 2.810 | 7.629 | 7.918 | -0.289 | -3.655 |
| Naphthalenedione | -534.682 | 1.114 | -6.151 | -9.659 | -4.171 | -1.671 | 1.979 | 7.988 | 8.403 | -0.415 | -4.935 |
| Nitrobenzene | -436.435 | -0.581 | -6.816 | -10.760 | -3.485 | -0.536 | 3.332 | 10.224 | 9.697 | 0.527 | 5.432 |
| Nitrobenzonitrile | -528.608 | 0.651 | -7.301 | -11.042 | -4.136 | -1.451 | 3.165 | 9.590 | 9.348 | 0.242 | 2.591 |
| Phenazine | -571.106 | 0.464 | -5.735 | -8.959 | -3.404 | -1.180 | 2.331 | 7.779 | 7.387 | 0.392 | 5.306 |
| Phthalicanhydride | -532.560 | -0.336 | -6.900 | -10.579 | -3.601 | -0.986 | 3.299 | 9.593 | 9.700 | -0.107 | -1.107 |
| Phthalimide | -512.703 | -0.964 | -6.350 | -9.972 | -3.299 | -0.741 | 3.051 | 9.231 | 9.482 | -0.251 | -2.650 |
| TCNE | -447.173 | 3.614 | -8.597 | -11.754 | -5.780 | -3.244 | 2.817 | 8.510 | 8.954 | -0.444 | -4.954 |
| TCNQ | -678.022 | 16.349 | -7.050 | -9.377 | -5.514 | -3.666 | 1.536 | 5.711 | 6.262 | -0.551 | -8.804 |
| Tetrachloro-isobenzofuranedione | -2370.300 | 3.504 | -7.106 | -9.777 | -4.062 | -1.721 | 3.044 | 8.057 | 8.430 | -0.373 | -4.427 |
| Tetrachlorobenzoquinone | -2218.912 | 3.724 | -6.983 | -9.787 | -5.010 | -2.604 | 1.973 | 7.184 | 7.762 | -0.578 | -7.451 |
| Tetrafluorobenzenedicarbonitrile | -813.119 | 1.210 | -7.444 | -10.070 | -4.133 | -1.703 | 3.311 | 8.367 | 9.151 | -0.784 | -8.569 |
| Tetrafluorobenzoquinone | -777.913 | 2.135 | -7.374 | -10.343 | -5.127 | -2.261 | 2.248 | 8.082 | 8.861 | -0.779 | -8.791 |
| mDCBN | -416.374 | 0.572 | -7.393 | -10.240 | -3.288 | -0.812 | 4.105 | 9.428 | 9.832 | -0.404 | -4.112 |
Table 7. Large organic molecules olLOSC calculation results with and . (See b and e in Figure 3 and Figure 4) Reference gap (Ref.) is CCSD-T benchmark from Ref. [43].
olLOSC beyond the random phase approximation
As we mentioned earlier, the RPA is an approximation; at first glance, it should appear that we should avoid it if possible. In the subsequent discourse, a comparative analysis will be conducted between the beyond RPA olLOSC scheme and the RPA-based olLOSC scheme. Owing to certain numerical instability, wherein the preconditioner employed in bulk calculations may lose its positive-definiteness when the potential term includes exchange and correlation effects, the beyond RPA olLOSC protocol demonstrates robustness exclusively in molecular systems. In Figure 5 and Figure 6, parameter scanning is performed analogously to the conventional RPA-based olLOSC. For two parameter sets with particular interest, and , we also provide their numerical results in Table 12. The beyond-RPA olLOSC routine yields superior results for small-sized molecules, yet underperforms in large-sized molecules, with comparable outcomes observed for both sets of parameters. Based on this comparison and on the basis of self-consistence and robustness, it is preferred to adapt the RPA-based olLOSC as a conventional olLOSC routine.

Figure 5. of-LOSC beyond RPA Relative errors (%) by LOSC and the vW in the molecules tested. (a, d): Small molecules. (b, e): Large molecules. (c, f): All molecular systems.

Figure 6. of-LOSC beyond RPA unscaled errors (eV) by LOSC and the vW in the molecules tested. (a, d): Small molecules. (b, e): Large molecules. (c, f): All molecular systems.
| γ | λ | isRPA | MAE (eV) | MSE (eV) | MARE | MSRE |
| Small Molecules (17 molecules) | ||||||
| 0.30 | 0.75 | TRUE | 0.562 | 0.528 | 5.479% | 5.077% |
| 0.30 | 0.75 | FALSE | 0.376 | 0.277 | 3.673% | 2.526% |
| 0.47714 | 1.0 | TRUE | 0.757 | 0.738 | 7.446% | 7.231% |
| 0.47714 | 1.0 | FALSE | 0.588 | 0.541 | 5.737% | 5.219% |
| Large Molecules (23 molecules) | ||||||
| 0.30 | 0.75 | TRUE | 0.352 | -0.251 | 4.176% | -3.017% |
| 0.30 | 0.75 | FALSE | 0.491 | -0.457 | 5.779% | -5.386% |
| 0.47714 | 1.0 | TRUE | 0.393 | -0.249 | 4.761% | -3.071% |
| 0.47714 | 1.0 | FALSE | 0.486 | -0.413 | 5.791% | -4.946% |
| All Molecules (40 molecules) | ||||||
| 0.30 | 0.75 | TRUE | 0.442 | 0.080 | 4.730% | 0.423% |
| 0.30 | 0.75 | FALSE | 0.443 | -0.145 | 4.884% | -2.024% |
| 0.47714 | 1.0 | TRUE | 0.548 | 0.171 | 5.902% | 1.307% |
| 0.47714 | 1.0 | FALSE | 0.529 | -0.008 | 5.768% | -0.626% |
Table 12. Comparison of the default olLOSC routine (within random phase approximation) with the olLOSC routine beyond random phase approximation across various categories of molecular systems. Also see Figure 3–Figure 6 for paramter set other than {, } and {, }. Raw data and visualization scripts are in the SI dataset.
Total energy
Materials
The materials we tested are in their equilibrium geometry, so we expect only a small correction from olLOSC to the total energy; and that is what we see. For the production dataset seen in the main text with (, ) (Fig. 1; Table 2), the maximum energy correction is Ry, for silicon; it also has the largest relative energy change, .
| System | k-mesh | Total energy (Ry) | VBM (eV) PBE | VBM (eV) olLOSC | CBM (eV) PBE | CBM (eV) olLOSC | Gap (eV) PBE | Gap (eV) olLOSC | Ref. | olLOSC gap error eV | olLOSC gap error % | |
| AlP | 06 | −18.58 | 5.021 | 4.359 | 6.597 | 6.709 | 1.576 | 2.350 | 2.606 | −0.256 | −9.808 | |
| Ar | 06 | −45.18 | −3.948 | −5.756 | 4.755 | 5.066 | 8.703 | 10.822 | 14.200 | −3.378 | −23.788 | |
| BN | 06 | −26.81 | 11.146 | 9.772 | 15.677 | 16.427 | 4.531 | 6.655 | 6.622 | 0.033 | 0.500 | |
| BP | 10 | −19.79 | 9.240 | 8.570 | 10.492 | 10.932 | 1.252 | 2.362 | 2.501 | −0.139 | −5.570 | |
| C | 08 | −24.08 | 13.355 | 12.820 | 17.544 | 18.235 | 4.189 | 5.415 | 5.803 | −0.388 | −6.693 | |
| Ge | 06 | −357.28 | 9.561 | 9.094 | 9.593 | 9.855 | 0.032 | 0.761 | 0.773 | −0.012 | −1.604 | |
| LiCl | 06 | −47.47 | 1.667 | 0.422 | 7.996 | 7.842 | 6.329 | 7.420 | 9.836 | −2.416 | −24.564 | |
| LiF | 06 | −63.98 | 0.970 | −1.702 | 10.156 | 10.241 | 9.187 | 11.943 | 15.431 | −3.488 | −22.605 | |
| MgO | 06 | −152.10 | 8.099 | 6.326 | 12.901 | 13.366 | 4.803 | 7.041 | 8.503 | −1.462 | −17.200 | |
| NaF | 06 | −140.97 | 0.999 | −1.639 | 7.389 | 8.002 | 6.390 | 9.641 | 12.199 | −2.558 | −20.971 | |
| Ne | 06 | −66.71 | −9.071 | −12.716 | 2.545 | 5.833 | 11.616 | 18.549 | 21.700 | −3.151 | −14.521 | |
| Si | 10 | −16.92 | 6.234 | 5.692 | 6.812 | 7.034 | 0.578 | 1.342 | 1.228 | 0.114 | 9.267 | |
| SiC | 08 | −20.54 | 9.496 | 8.635 | 10.857 | 11.051 | 1.361 | 2.416 | 2.595 | −0.179 | −6.913 |
Table 2. The dataset seen in Fig. 1 (b) of the main text: the olLOSC parameters are , We use a Monkhorst–Pack -mesh, except in cases where the material’s conduction band minimum does not lie on a high-symmetry -point (as evidenced by the PBE CBM changing with ). Thus, we use -points for C and SiC, and -points for BP and Si.
Molecules
Total energies in the underlying DFA calculations for molecules are not size-consistent [22], so a large is expected for large molecules. Table 7 and 8 include and for the large molecules we tested. We observed olLOSC energy corrections up to a few mE. The largest is for TCNQ (tetracyanoquinodimethane), which is a highly conjugated molecule (thus, significant delocalization error is expected). It has 16.349 when and . Note that increased spatial localization (a smaller ) yields a larger because non-integer local occupations are penalized less. Thus, when and , for TCNQ is 11.383 . The reaction barrier height is heavily based on total energy calculations. Table 13 includes mean signed error (MSE) and mean absolute error (MAE) for reaction barrier heights using the HTBH38 and NHTBH38 datasets [26, 27], where reference reaction barrier heights are considered to be the ’REF1’ (exclude relativistic effect) within these datasets. A value of 0.47714 generally replicates the results of PBE reaction barrier heights, whereas a value of 0.3 yields the lowest mean absolute error.
| NHTBH38 | HTBH38 | All reactions | |||||
| MAE | MSE | MAE | MSE | MAE | MSE | ||
| PBE | N/A | 9.915 | -8.141 | 9.424 | -9.424 | 9.669 | -8.782 |
| 0.2 | 0.75 | 7.888 | -2.721 | 9.873 | -9.873 | 8.881 | -6.297 |
| 0.3 | 0.75 | 7.012 | -4.937 | 10.116 | -10.116 | 8.564 | -7.527 |
| 0.47714 | 0.75 | 9.249 | -7.438 | 9.788 | -9.788 | 9.518 | -8.613 |
Table 13. Mean signed error (MSE) and mean absolute error (MAE) for reaction barrier heights (in kcal/mol). NHTBH38 is a 38-reaction dataset of non-hydrogen transfer reaction barrier heights, while HTBH38 contains 38 hydrogen transfer barrier heights. We take REF1 (which excludes relativistic effect) from [26, 27] as a reference. The first row gives the reaction barrier calculated by the parent DFT method (PBE); the subsequent rows detail the olLOSC reaction barrier calculations using 75% von Weizsäcker kinetic energy () at different localization parameters .
Local minima in dually localized Wannier functions
DLWFs [3] are computed variationally, minimizing for each spin component the cost function
where and . (Note that , so we do not need to sum over the unit cell index ; that would only multiply by a constant, which does not affect the optimization.)
We minimize by modifying the method used for maximally localized Wannier functions (MLWFs) [4]. The difference is in the cost function, which includes the Hamiltonian, as first suggested by [28]. The key difference is that DLWFs are constructed from both valence and conduction bands. We first disentangle the conduction’ manifold [29], obtaining optimally smooth pseudo-bands that span the virtual subspace. (If the disentangled bands are analytic, the MLWFs they specify decay exponentially [30].) Once the bands are chosen, we formulate as a function of the elements of the localization unitaries , which localize the disentangled bands:
Then we minimize with the conjugate gradient method.
It was noted as early as [4] that, depending on the quality of the initial guess, the MLWF localization scheme can become trapped in local minima of the cost function. More than two decades of work has rendered MLWFs extremely robust to local minima, to the point that there are automated and high-throughput workflows for Wannierization [31]. This scheme scheme relies on the selected columns of the density matrix (SCDM) algorithm [32, 33] for its initial guess. SCDM produces orthonormal, localized Wannier-like functions deterministically by exploiting the exponential decay of the one-electron reduced density matrix [34],
We, too, use SCDM to generate initial guesses for DLWFs, but our minimization is still sensitive to local minima. This is due largely to the disentangled conduction bands, which have large energy variance . We avoid this as best we can by minimizing several times, over a variety of conjugate gradient step sizes, and choosing the smallest cost function value overall.
Mahler et al. also observed that the qualitative nature of the DLWFs changes as a function of (see Fig. 1 of [3]). At (the MLWF limit), all the WFs are degenerate; they are also typically equivalent up to symmetry operations in the primitive cell. At small positive , the valence and conduction DLWFs split off from one another. They have different shapes and different average energies . Splitting of degenerate subsets continues as is increased further.
We see similar patterns in the DLWFs for these systems, but the degeneracy pattern varies somewhat with the -sampling—especially for our largest , 0.47714. As a particularly noticeable example, observe the large difference (0.442 eV) in the olLOSC gap (, ) of BP between k = 6 and k = 8 (Figure 2(d) and Table 3). We attribute this difference to the DLWFs. Table 6 demonstrates how BP has two ‘classes’ of occupied DLWFs (with local occupations ) when computed with the 6 × 6 × 6 k-mesh: one has , and three are nearly degenerate, with . In the -sampled calculation, the lowest-energy DLWF is qualitatively very similar, but the other three occupied DLWFs have split into two sub-bands. One has , and two are nearly degenerate, with . They also have qualitatively different spatial spreads, as seen by the values of in Table 6.
Table (continued in Table 4): Choosing the and parameters in bulk systems (see Fig. 1). The k-mesh is as in Table 2. Reference gap (Ref.) is experimental − ZPR from Table 1.
| System | DLWF | vW | VBM PBE (eV) | VBM olLOSC (eV) | CBM PBE (eV) | CBM olLOSC (eV) | Gap PBE (eV) | Gap olLOSC (eV) | Ref. | eV | % | |
| AlP | 0.20 | 0.00 | 5.021 | 4.416 | 6.597 | 6.482 | 1.576 | 2.066 | 2.606 | −0.540 | −20.733 | |
| AlP | 0.20 | 0.75 | 5.021 | 4.349 | 6.597 | 6.710 | 1.576 | 2.361 | 2.606 | −0.245 | −9.405 | |
| AlP | 0.20 | 1.00 | 5.021 | 4.335 | 6.597 | 6.732 | 1.576 | 2.397 | 2.606 | −0.209 | −8.035 | |
| AlP | 0.30 | 0.00 | 5.021 | 4.425 | 6.597 | 6.489 | 1.576 | 2.064 | 2.606 | −0.542 | −20.814 | |
| AlP | 0.30 | 0.75 | 5.021 | 4.359 | 6.597 | 6.709 | 1.576 | 2.350 | 2.606 | −0.256 | −9.808 | |
| AlP | 0.30 | 1.00 | 5.021 | 4.345 | 6.597 | 6.730 | 1.576 | 2.385 | 2.606 | −0.221 | −8.480 | |
| AlP | 0.48 | 0.00 | 5.021 | 4.438 | 6.597 | 6.501 | 1.576 | 2.063 | 2.606 | −0.543 | −20.837 | |
| AlP | 0.48 | 0.75 | 5.021 | 4.373 | 6.597 | 6.713 | 1.576 | 2.340 | 2.606 | −0.266 | −10.207 | |
| AlP | 0.48 | 1.00 | 5.021 | 4.360 | 6.597 | 6.733 | 1.576 | 2.373 | 2.606 | −0.232 | −8.922 | |
| Ar | 0.20 | 0.00 | −3.948 | −9.236 | 4.755 | −9.282 | 8.703 | −0.047 | 14.200 | −14.247 | −100.329 | |
| Ar | 0.20 | 0.75 | −3.948 | −5.761 | 4.755 | 4.582 | 8.703 | 10.343 | 14.200 | −3.857 | −27.162 | |
| Ar | 0.20 | 1.00 | −3.948 | −5.778 | 4.755 | 4.645 | 8.703 | 10.423 | 14.200 | −3.777 | −26.599 | |
| Ar | 0.30 | 0.00 | −3.948 | −6.830 | 4.755 | −7.730 | 8.703 | −0.900 | 14.200 | −15.100 | −106.339 | |
| Ar | 0.30 | 0.75 | −3.948 | −5.756 | 4.755 | 5.066 | 8.703 | 10.822 | 14.200 | −3.378 | −23.788 | |
| Ar | 0.30 | 1.00 | −3.948 | −5.774 | 4.755 | 5.119 | 8.703 | 10.893 | 14.200 | −3.307 | −23.291 | |
| Ar | 0.48 | 0.00 | −3.948 | −5.601 | 4.755 | −6.469 | 8.703 | −0.868 | 14.200 | −15.068 | −106.115 | |
| Ar | 0.48 | 0.75 | −3.948 | −5.752 | 4.755 | 5.109 | 8.703 | 10.861 | 14.200 | −3.339 | −23.513 | |
| Ar | 0.48 | 1.00 | −3.948 | −5.769 | 4.755 | 5.158 | 8.703 | 10.928 | 14.200 | −3.272 | −23.044 | |
| BN | 0.20 | 0.00 | 11.146 | 9.860 | 15.677 | 16.548 | 4.531 | 6.689 | 6.622 | 0.067 | 1.010 | |
| BN | 0.20 | 0.75 | 11.146 | 9.764 | 15.677 | 16.745 | 4.531 | 6.981 | 6.622 | 0.359 | 5.423 | |
| BN | 0.20 | 1.00 | 11.146 | 9.743 | 15.677 | 16.781 | 4.531 | 7.037 | 6.622 | 0.415 | 6.275 | |
| BN | 0.30 | 0.00 | 11.146 | 9.867 | 15.677 | 16.250 | 4.531 | 6.383 | 6.622 | −0.239 | −3.617 | |
| BN | 0.30 | 0.75 | 11.146 | 9.772 | 15.677 | 16.427 | 4.531 | 6.655 | 6.622 | 0.033 | 0.500 | |
| BN | 0.30 | 1.00 | 11.146 | 9.752 | 15.677 | 16.458 | 4.531 | 6.706 | 6.622 | 0.084 | 1.272 | |
| BN | 0.48 | 0.00 | 11.146 | 10.372 | 15.677 | 16.311 | 4.531 | 5.939 | 6.622 | −0.683 | −10.307 | |
| BN | 0.48 | 0.75 | 11.146 | 10.333 | 15.677 | 16.400 | 4.531 | 6.067 | 6.622 | −0.555 | −8.380 | |
| BN | 0.48 | 1.00 | 11.146 | 10.325 | 15.677 | 16.418 | 4.531 | 6.093 | 6.622 | −0.529 | −7.982 | |
| BP | 0.20 | 0.00 | 9.240 | 8.629 | 10.492 | 10.850 | 1.252 | 2.221 | 2.501 | −0.280 | −11.188 | |
| BP | 0.20 | 0.75 | 9.240 | 8.556 | 10.492 | 10.935 | 1.252 | 2.378 | 2.501 | −0.123 | −4.910 | |
| BP | 0.20 | 1.00 | 9.240 | 8.540 | 10.492 | 10.949 | 1.252 | 2.409 | 2.501 | −0.092 | −3.675 | |
| BP | 0.30 | 0.00 | 9.240 | 8.641 | 10.492 | 10.848 | 1.252 | 2.208 | 2.501 | −0.293 | −11.731 | |
| BP | 0.30 | 0.75 | 9.240 | 8.570 | 10.492 | 10.932 | 1.252 | 2.362 | 2.501 | −0.139 | −5.570 | |
| BP | 0.30 | 1.00 | 9.240 | 8.554 | 10.492 | 10.946 | 1.252 | 2.392 | 2.501 | −0.109 | −4.362 | |
| BP | 0.48 | 0.00 | 9.240 | 8.591 | 10.492 | 10.819 | 1.252 | 2.227 | 2.501 | −0.274 | −10.944 | |
| BP | 0.48 | 0.75 | 9.240 | 8.576 | 10.492 | 10.827 | 1.252 | 2.251 | 2.501 | −0.250 | −9.996 | |
| C | 0.20 | 0.00 | 13.355 | 12.390 | 17.544 | 18.504 | 4.189 | 6.114 | 5.803 | 0.311 | 5.352 | |
| C | 0.20 | 0.75 | 13.355 | 12.303 | 17.544 | 18.638 | 4.189 | 6.335 | 5.803 | 0.532 | 9.164 | |
| C | 0.20 | 1.00 | 13.355 | 12.283 | 17.544 | 18.664 | 4.189 | 6.381 | 5.803 | 0.578 | 9.962 | |
| C | 0.30 | 0.00 | 13.355 | 12.852 | 17.544 | 18.158 | 4.189 | 5.307 | 5.803 | −0.496 | −8.554 | |
| C | 0.30 | 0.75 | 13.355 | 12.820 | 17.544 | 18.235 | 4.189 | 5.415 | 5.803 | −0.388 | −6.693 | |
| C | 0.30 | 1.00 | 13.355 | 12.814 | 17.544 | 18.249 | 4.189 | 5.435 | 5.803 | −0.368 | −6.333 | |
| C | 0.48 | 0.00 | 13.355 | 12.933 | 17.544 | 18.034 | 4.189 | 5.101 | 5.803 | −0.702 | −12.104 | |
| C | 0.48 | 0.75 | 13.355 | 12.911 | 17.544 | 18.086 | 4.189 | 5.175 | 5.803 | −0.628 | −10.815 | |
| C | 0.48 | 1.00 | 13.355 | 12.906 | 17.544 | 18.096 | 4.189 | 5.190 | 5.803 | −0.613 | −10.569 | |
| Ge | 0.20 | 0.00 | 9.561 | 9.192 | 9.593 | 9.748 | 0.032 | 0.555 | 0.773 | −0.218 | −28.150 | |
| Ge | 0.20 | 0.75 | 9.561 | 9.060 | 9.593 | 9.855 | 0.032 | 0.795 | 0.773 | 0.022 | 2.885 | |
| Ge | 0.20 | 1.00 | 9.561 | 9.031 | 9.593 | 9.874 | 0.032 | 0.843 | 0.773 | 0.070 | 9.056 | |
| Ge | 0.30 | 0.00 | 9.561 | 9.216 | 9.593 | 9.747 | 0.032 | 0.530 | 0.773 | −0.243 | −31.384 | |
| Ge | 0.30 | 0.75 | 9.561 | 9.094 | 9.593 | 9.855 | 0.032 | 0.761 | 0.773 | −0.012 | −1.604 | |
| Ge | 0.30 | 1.00 | 9.561 | 9.067 | 9.593 | 9.873 | 0.032 | 0.806 | 0.773 | 0.033 | 4.295 | |
| Ge | 0.48 | 0.00 | 9.561 | 9.359 | 9.593 | 9.735 | 0.032 | 0.377 | 0.773 | −0.396 | −51.294 | |
| Ge | 0.48 | 0.75 | 9.561 | 9.288 | 9.593 | 9.842 | 0.032 | 0.553 | 0.773 | −0.220 | −28.396 | |
| Ge | 0.48 | 1.00 | 9.561 | 9.273 | 9.593 | 9.860 | 0.032 | 0.587 | 0.773 | −0.186 | −24.036 | |
| System | VBM PBE (eV) | VBM olLOSC (eV) | CBM PBE (eV) | CBM olLOSC (eV) | Gap PBE (eV) | Gap olLOSC (eV) | Ref. | olLOSC gap error (eV) | olLOSC gap error (%) | |||
| LiCl | 0.20 | 0.00 | 1.667 | 0.538 | 7.996 | 4.990 | 6.329 | 4.452 | 9.836 | −5.384 | −54.737 | |
| LiCl | 0.20 | 0.75 | 1.667 | 0.420 | 7.996 | 7.749 | 6.329 | 7.329 | 9.836 | −2.507 | −25.491 | |
| LiCl | 0.20 | 1.00 | 1.667 | 0.404 | 7.996 | 7.800 | 6.329 | 7.396 | 9.836 | −2.441 | −24.812 | |
| LiCl | 0.30 | 0.00 | 1.667 | 0.540 | 7.996 | 5.522 | 6.329 | 4.982 | 9.836 | −4.854 | −49.349 | |
| LiCl | 0.30 | 0.75 | 1.667 | 0.422 | 7.996 | 7.842 | 6.329 | 7.420 | 9.836 | −2.416 | −24.564 | |
| LiCl | 0.30 | 1.00 | 1.667 | 0.406 | 7.996 | 7.886 | 6.329 | 7.480 | 9.836 | −2.356 | −23.952 | |
| LiCl | 0.48 | 0.00 | 1.667 | 0.545 | 7.996 | 5.959 | 6.329 | 5.414 | 9.836 | −4.422 | −44.957 | |
| LiCl | 0.48 | 0.75 | 1.667 | 0.427 | 7.996 | 7.940 | 6.329 | 7.513 | 9.836 | −2.323 | −23.618 | |
| LiCl | 0.48 | 1.00 | 1.667 | 0.411 | 7.996 | 7.978 | 6.329 | 7.567 | 9.836 | −2.269 | −23.072 | |
| LiF | 0.20 | 0.00 | 0.970 | −1.543 | 10.156 | 7.622 | 9.187 | 9.165 | 15.431 | −6.265 | −40.603 | |
| LiF | 0.20 | 0.75 | 0.970 | −1.711 | 10.156 | 10.167 | 9.187 | 11.878 | 15.431 | −3.553 | −23.024 | |
| LiF | 0.20 | 1.00 | 0.970 | −1.738 | 10.156 | 10.225 | 9.187 | 11.963 | 15.431 | −3.468 | −22.472 | |
| LiF | 0.30 | 0.00 | 0.970 | −1.534 | 10.156 | 7.911 | 9.187 | 9.445 | 15.431 | −5.986 | −38.795 | |
| LiF | 0.30 | 0.75 | 0.970 | −1.702 | 10.156 | 10.241 | 9.187 | 11.943 | 15.431 | −3.488 | −22.605 | |
| LiF | 0.30 | 1.00 | 0.970 | −1.729 | 10.156 | 10.295 | 9.187 | 12.024 | 15.431 | −3.407 | −22.079 | |
| LiF | 0.48 | 0.00 | 0.970 | −1.526 | 10.156 | 8.257 | 9.187 | 9.783 | 15.431 | −5.648 | −36.605 | |
| LiF | 0.48 | 0.75 | 0.970 | −1.695 | 10.156 | 10.330 | 9.187 | 12.025 | 15.431 | −3.406 | −22.076 | |
| LiF | 0.48 | 1.00 | 0.970 | −1.722 | 10.156 | 10.379 | 9.187 | 12.100 | 15.431 | −3.331 | −21.584 | |
| MgO | 0.20 | 0.00 | 8.099 | 6.446 | 12.901 | 12.508 | 4.803 | 6.061 | 8.503 | −2.442 | −28.716 | |
| MgO | 0.20 | 0.75 | 8.099 | 6.312 | 12.901 | 13.246 | 4.803 | 6.934 | 8.503 | −1.569 | −18.453 | |
| MgO | 0.20 | 1.00 | 8.099 | 6.288 | 12.901 | 13.282 | 4.803 | 6.994 | 8.503 | −1.509 | −17.748 | |
| MgO | 0.30 | 0.00 | 8.099 | 6.461 | 12.901 | 12.905 | 4.803 | 6.444 | 8.503 | −2.059 | −24.209 | |
| MgO | 0.30 | 0.75 | 8.099 | 6.326 | 12.901 | 13.366 | 4.803 | 7.041 | 8.503 | −1.462 | −17.200 | |
| MgO | 0.30 | 1.00 | 8.099 | 6.302 | 12.901 | 13.394 | 4.803 | 7.092 | 8.503 | −1.411 | −16.592 | |
| MgO | 0.48 | 0.00 | 8.099 | 6.470 | 12.901 | 13.090 | 4.803 | 6.620 | 8.503 | −1.883 | −22.146 | |
| MgO | 0.48 | 0.75 | 8.099 | 6.335 | 12.901 | 13.409 | 4.803 | 7.074 | 8.503 | −1.429 | −16.811 | |
| MgO | 0.48 | 1.00 | 8.099 | 6.311 | 12.901 | 13.428 | 4.803 | 7.117 | 8.503 | −1.386 | −16.304 | |
| NaF | 0.20 | 0.00 | 0.999 | −1.485 | 7.389 | 6.396 | 6.390 | 7.881 | 12.199 | −4.318 | −35.396 | |
| NaF | 0.20 | 0.75 | 0.999 | −1.652 | 7.389 | 7.999 | 6.390 | 9.652 | 12.199 | −2.547 | −20.883 | |
| NaF | 0.20 | 1.00 | 0.999 | −1.679 | 7.389 | 8.035 | 6.390 | 9.714 | 12.199 | −2.485 | −20.370 | |
| NaF | 0.30 | 0.00 | 0.999 | −1.470 | 7.389 | 6.540 | 6.390 | 8.010 | 12.199 | −4.189 | −34.337 | |
| NaF | 0.30 | 0.75 | 0.999 | −1.639 | 7.389 | 8.002 | 6.390 | 9.641 | 12.199 | −2.558 | −20.971 | |
| NaF | 0.30 | 1.00 | 0.999 | −1.666 | 7.389 | 8.035 | 6.390 | 9.700 | 12.199 | −2.499 | −20.482 | |
| NaF | 0.48 | 0.00 | 0.999 | −1.458 | 7.389 | 6.743 | 6.390 | 8.201 | 12.199 | −3.998 | −32.772 | |
| NaF | 0.48 | 0.75 | 0.999 | −1.627 | 7.389 | 8.005 | 6.390 | 9.632 | 12.199 | −2.567 | −21.039 | |
| NaF | 0.48 | 1.00 | 0.999 | −1.654 | 7.389 | 8.034 | 6.390 | 9.688 | 12.199 | −2.511 | −20.586 | |
| Ne | 0.20 | 0.00 | −9.071 | −104.215 | 2.545 | −109.062 | 11.616 | −4.847 | 21.700 | −26.547 | −122.336 | |
| Ne | 0.20 | 0.75 | −9.071 | −12.721 | 2.545 | 6.016 | 11.616 | 18.737 | 21.700 | −2.963 | −13.654 | |
| Ne | 0.20 | 1.00 | −9.071 | −12.756 | 2.545 | 6.177 | 11.616 | 18.933 | 21.700 | −2.767 | −12.749 | |
| Ne | 0.30 | 0.00 | −9.071 | −109.389 | 2.545 | −96.808 | 11.616 | 12.581 | 21.700 | −9.119 | −42.021 | |
| Ne | 0.30 | 0.75 | −9.071 | −12.716 | 2.545 | 5.833 | 11.616 | 18.549 | 21.700 | −3.151 | −14.521 | |
| Ne | 0.30 | 1.00 | −9.071 | −12.751 | 2.545 | 5.980 | 11.616 | 18.732 | 21.700 | −2.968 | −13.680 | |
| Ne | 0.48 | 0.00 | −9.071 | −74.744 | 2.545 | −36.434 | 11.616 | 38.310 | 21.700 | 16.610 | 76.544 | |
| Ne | 0.48 | 0.75 | −9.071 | −12.712 | 2.545 | 5.619 | 11.616 | 18.330 | 21.700 | −3.370 | −15.529 | |
| Ne | 0.48 | 1.00 | −9.071 | −12.747 | 2.545 | 5.751 | 11.616 | 18.499 | 21.700 | −3.201 | −14.753 | |
| Si | 0.20 | 0.00 | 6.234 | 5.691 | 6.812 | 6.973 | 0.578 | 1.282 | 1.228 | 0.054 | 4.406 | |
| Si | 0.20 | 0.75 | 6.234 | 5.618 | 6.812 | 7.034 | 0.578 | 1.416 | 1.228 | 0.188 | 15.309 | |
| Si | 0.20 | 1.00 | 6.234 | 5.601 | 6.812 | 7.042 | 0.578 | 1.441 | 1.228 | 0.213 | 17.386 | |
| Si | 0.30 | 0.00 | 6.234 | 5.755 | 6.812 | 6.973 | 0.578 | 1.218 | 1.228 | −0.010 | −0.790 | |
| Si | 0.30 | 0.75 | 6.234 | 5.692 | 6.812 | 7.034 | 0.578 | 1.342 | 1.228 | 0.114 | 9.267 | |
| Si | 0.30 | 1.00 | 6.234 | 5.677 | 6.812 | 7.042 | 0.578 | 1.365 | 1.228 | 0.137 | 11.132 | |
| Si | 0.48 | 0.75 | 6.234 | 5.735 | 6.812 | 7.033 | 0.578 | 1.298 | 1.228 | 0.070 | 5.692 | |
| Si | 0.48 | 1.00 | 6.234 | 5.722 | 6.812 | 7.041 | 0.578 | 1.319 | 1.228 | 0.091 | 7.427 | |
| SiC | 0.20 | 0.00 | 9.496 | 8.690 | 10.857 | 10.961 | 1.361 | 2.271 | 2.595 | −0.324 | −12.470 | |
| SiC | 0.20 | 0.75 | 9.496 | 8.617 | 10.857 | 11.184 | 1.361 | 2.567 | 2.595 | −0.028 | −1.094 | |
| SiC | 0.20 | 1.00 | 9.496 | 8.601 | 10.857 | 11.216 | 1.361 | 2.615 | 2.595 | 0.020 | 0.759 | |
| SiC | 0.30 | 0.00 | 9.496 | 8.707 | 10.857 | 10.902 | 1.361 | 2.195 | 2.595 | −0.400 | −15.399 | |
| SiC | 0.30 | 0.75 | 9.496 | 8.635 | 10.857 | 11.051 | 1.361 | 2.416 | 2.595 | −0.179 | −6.913 | |
| SiC | 0.30 | 1.00 | 9.496 | 8.619 | 10.857 | 11.069 | 1.361 | 2.450 | 2.595 | −0.145 | −5.580 | |
| SiC | 0.48 | 0.00 | 9.496 | 8.863 | 10.857 | 10.894 | 1.361 | 2.031 | 2.595 | −0.564 | −21.738 | |
| SiC | 0.48 | 0.75 | 9.496 | 8.817 | 10.857 | 11.008 | 1.361 | 2.191 | 2.595 | −0.404 | −15.584 | |
| SiC | 0.48 | 1.00 | 9.496 | 8.807 | 10.857 | 11.021 | 1.361 | 2.214 | 2.595 | −0.381 | −14.678 |
Table 3. This table (continued in Table 4): Choosing the and parameters in bulk systems (see Fig. 1). The k-mesh is as in Table 2. Reference gap (Ref.) is experimental − ZPR from Table 1.
| Local occupation | Average energy (eV) | Spatial spread () | |||
| 1.0000 | 1.0000 | −3.075373 | −3.065085 | 1.827005 | 1.840974 |
| 0.9995 | 0.9997 | 4.702881 | 3.041265 | 2.313304 | 3.948855 |
| 0.9995 | 0.9988 | 4.703143 | 5.511713 | 2.313211 | 3.489266 |
| 0.9995 | 0.9988 | 4.703387 | 5.532760 | 2.313190 | 3.553196 |
| 0.0008 | 0.0014 | 14.526515 | 14.583134 | 4.215674 | 4.342439 |
| 0.0004 | 0.0008 | 15.866122 | 15.754140 | 4.308766 | 4.357879 |
| 0.0002 | 0.0004 | 16.851640 | 16.735331 | 3.598073 | 3.649210 |
| 0.0000 | 0.0001 | 18.791317 | 18.544373 | 4.529317 | 4.627405 |
Table 6. The DLWFs () for BP, arranged in order of increasing energy. The local occupations are the diagonal elements of the density matrix in the DLWF basis, ; they are real, with , so they measure DLWF occupation. Observe the threefold near-degeneracy of the occupied DLWFs in the -sampled calculation and how it differs from the occupied submanifold in the calculation.
Differences in spatial localization change, especially, the Coulomb interaction in the curvature, leading to a different correction to the band gap. In particular, observe from Figure 2 (d) that the calculation has a substantially higher gap than its -sampled counterpart, corresponding to the increased spatial localization (smaller ) of its occupied DLWFs.
We are currently seeking an alternative to the DLWF minimization that avoids this weakness. Since there is no physical principle that specifies what should be, we believe that a deterministic algorithm—yielding Wannier functions approximately localized in both space and energy—will improve the robustness of olLOSC.
Figure 1 (PDF p. 37). (a) Mean absolute error (%) for small-gap ( eV) systems. (b) Mean absolute error (%) for all systems. (c) Mean signed error (%) for small-gap ( eV) systems. (d) Mean signed error (%) for all systems. (e) Mean absolute error (eV) for small-gap ( eV) systems. (f) Mean absolute error (eV) for all systems. (g) Mean signed error (eV) for small-gap ( eV) systems. (h) Mean signed error (eV) for all systems. Figure 1: Errors by DLWF mixing and the vW in the materials tested. (a–d): Relative errors (%); (e–h): unscaled errors (eV). (a, c, e, g): Small-gapped systems (gaps eV), for which olLOSC is most accurate; (b, d, f, h): All systems. (a, b, e, f): Absolute values; (c, d, g, h): Signed errors.
| System | Lattice | (Å) | Gap (eV) | ZPR (eV) | Gap − ZPR (eV) |
| C | A4 | 3.57 | 5.48 | −0.32 [40] | 5.80 |
| Si | A4 | 5.43 | 1.17 | −0.06 [40] | 1.23 |
| SiC | B3 | 4.36 | 2.42 | −0.17 [40] | 2.59 |
| Ge | A4 | 5.66 | 0.74 | −0.03 [41] | 0.77 |
| MgO | B1 | 4.21 | 7.97 | −0.53 [40] | 7.75 |
| LiF | B1 | 4.02 | 14.20 | −1.23 [40] | 15.43 |
| LiCl | B1 | 5.13 | 9.40 | −0.44 [42] | 9.84 |
| NaF | B1 | 4.62 | 11.50 | −0.70 [42] | 12.20 |
| BN | B3 | 3.62 | 6.22 | −0.40 [40] | 6.62 |
| BP | B3 | 4.54 | 2.40 | −0.10 [42] | 2.50 |
| AlP | B3 | 5.46 | 2.51 | −0.10 [40] | 2.61 |
| Ne | A1 | 4.43 | 21.70 | n/a | 21.70 |
| Ar | A1 | 5.26 | 14.20 | n/a | 14.20 |
Table 1 (PDF p. 44). The bulk systems studied. The lattice is defined by the Strukturbericht symbol (A4: diamond; B3: zincblende; B1: halite; B3: wurtzite; A1: simple cubic). is the experimental lattice parameter, in angstrom; these are taken from [35] when available, [36] otherwise. The gap is experimental, taken from [37] (LiF, LiCl, Ne, Ar); [38] (NaF); [39 p. 184] (MgO); and [35] and containing references (the rest). ZPR is zero-point renormalization, computed in the indicated reference for each system.
| System | k-mesh | VBM PBE (eV) | VBM olLOSC (eV) | CBM PBE (eV) | CBM olLOSC (eV) | Gap PBE (eV) | Gap olLOSC (eV) | Ref. | olLOSC gap error eV | % | ||||
| AlP | 06 | 0.30 | 0.75 | −18.58 | 5.021 | 4.150 | 6.597 | 6.816 | 1.576 | 2.666 | 2.606 | 0.060 | 2.306 | |
| AlP | 08 | 0.30 | 0.75 | −18.58 | 5.016 | 4.172 | 6.595 | 6.792 | 1.579 | 2.620 | 2.606 | 0.014 | 0.553 | |
| AlP | 06 | 0.30 | 1.00 | −18.58 | 5.021 | 4.099 | 6.597 | 6.856 | 1.576 | 2.757 | 2.606 | 0.151 | 5.810 | |
| Ar | 06 | 0.30 | 0.75 | −45.18 | −3.948 | −5.909 | 4.755 | 5.079 | 8.703 | 10.988 | 14.200 | −3.212 | −22.620 | |
| Ar | 08 | 0.30 | 0.75 | −45.18 | −3.948 | −5.759 | 4.755 | 4.760 | 8.703 | 10.519 | 14.200 | −3.681 | −25.923 | |
| Ar | 06 | 0.30 | 1.00 | −45.18 | −3.948 | −6.034 | 4.755 | 5.154 | 8.703 | 11.188 | 14.200 | −3.012 | −21.213 | |
| BN | 06 | 0.30 | 0.75 | −26.81 | 11.146 | 9.716 | 15.677 | 16.495 | 4.531 | 6.778 | 6.622 | 0.156 | 2.362 | |
| BN | 08 | 0.30 | 0.75 | −26.81 | 11.143 | 9.800 | 15.676 | 16.775 | 4.533 | 6.975 | 6.622 | 0.353 | 5.331 | |
| BN | 06 | 0.30 | 1.00 | −26.81 | 11.146 | 9.641 | 15.677 | 16.560 | 4.531 | 6.920 | 6.622 | 0.298 | 4.494 | |
| BP | 06 | 0.30 | 0.75 | −19.79 | 9.246 | 8.290 | 10.591 | 11.268 | 1.345 | 2.978 | 2.501 | 0.477 | 19.060 | |
| BP | 08 | 0.30 | 0.75 | −19.79 | 9.241 | 8.316 | 10.522 | 11.158 | 1.281 | 2.842 | 2.501 | 0.341 | 13.643 | |
| BP | 06 | 0.30 | 1.00 | −19.79 | 9.246 | 8.215 | 10.591 | 11.332 | 1.345 | 3.117 | 2.501 | 0.616 | 24.630 | |
| C | 06 | 0.30 | 0.75 | −24.07 | 13.358 | 12.785 | 17.563 | 18.289 | 4.205 | 5.505 | 5.803 | −0.299 | −5.144 | |
| C | 08 | 0.30 | 0.75 | −24.08 | 13.355 | 12.825 | 17.544 | 18.224 | 4.189 | 5.399 | 5.803 | −0.404 | −6.962 | |
| C | 06 | 0.30 | 1.00 | −24.07 | 13.358 | 12.778 | 17.563 | 18.304 | 4.205 | 5.526 | 5.803 | −0.277 | −4.770 | |
| Ge | 06 | 0.30 | 0.75 | −357.28 | 9.561 | 8.932 | 9.593 | 9.975 | 0.032 | 1.043 | 0.773 | 0.270 | 34.955 | |
| Ge | 08 | 0.30 | 0.75 | −357.28 | 9.554 | 8.933 | 9.585 | 9.951 | 0.031 | 1.018 | 0.773 | 0.245 | 31.759 | |
| Ge | 06 | 0.30 | 1.00 | −357.28 | 9.561 | 8.866 | 9.593 | 10.032 | 0.032 | 1.165 | 0.773 | 0.392 | 50.776 | |
| LiCl | 06 | 0.30 | 0.75 | −47.47 | 1.667 | 0.228 | 7.996 | 7.892 | 6.329 | 7.664 | 9.836 | −2.172 | −22.085 | |
| LiCl | 08 | 0.30 | 0.75 | −47.47 | 1.667 | 0.303 | 7.996 | 7.427 | 6.329 | 7.124 | 9.836 | −2.712 | −27.573 | |
| LiCl | 06 | 0.30 | 1.00 | −47.47 | 1.667 | 0.109 | 7.996 | 7.988 | 6.329 | 7.879 | 9.836 | −1.957 | −19.899 | |
| LiF | 06 | 0.30 | 0.75 | −63.98 | 0.970 | −1.702 | 10.156 | 10.241 | 9.187 | 11.943 | 15.431 | −3.488 | −22.604 | |
| LiF | 08 | 0.30 | 0.75 | −63.98 | 0.970 | −1.539 | 10.156 | 10.088 | 9.187 | 11.626 | 15.431 | −3.805 | −24.658 | |
| LiF | 06 | 0.30 | 1.00 | −63.98 | 0.970 | −1.728 | 10.156 | 10.296 | 9.187 | 12.024 | 15.431 | −3.407 | −22.077 | |
| MgO | 06 | 0.30 | 0.75 | −152.10 | 8.099 | 6.426 | 12.901 | 13.362 | 4.803 | 6.936 | 8.503 | −1.567 | −19.257 | |
| MgO | 08 | 0.30 | 0.75 | −152.10 | 8.099 | 6.442 | 12.901 | 13.389 | 4.803 | 6.947 | 8.503 | −1.556 | −18.304 | |
| MgO | 06 | 0.30 | 1.00 | −152.10 | 8.099 | 6.302 | 12.901 | 13.369 | 4.803 | 7.067 | 8.503 | −1.436 | −16.897 | |
| Na | 06 | 0.30 | 0.75 | −140.17 | −9.071 | −12.638 | 2.545 | 5.796 | 11.616 | 18.434 | 21.700 | −3.266 | −15.050 | |
| Na | 08 | 0.30 | 0.75 | −140.17 | −9.071 | −12.483 | 2.417 | 5.384 | 11.488 | 17.867 | 21.700 | −3.833 | −17.664 | |
| Na | 06 | 0.30 | 1.00 | −140.17 | −9.071 | −12.680 | 2.545 | 5.796 | 11.616 | 18.476 | 21.700 | −3.224 | −14.853 | |
| Si | 06 | 0.30 | 0.75 | −16.92 | 6.235 | 5.492 | 6.847 | 7.172 | 0.612 | 1.680 | 1.228 | 0.452 | 36.816 | |
| Si | 08 | 0.30 | 0.75 | −16.92 | 6.235 | 5.544 | 6.847 | 7.172 | 0.612 | 1.728 | 1.228 | 0.500 | 40.713 | |
| Si | 06 | 0.30 | 1.00 | −16.92 | 6.235 | 5.441 | 6.847 | 7.313 | 0.612 | 1.872 | 1.228 | 0.644 | 52.443 | |
| SiC | 06 | 0.30 | 0.75 | −20.54 | 9.463 | 8.408 | 10.857 | 11.123 | 1.361 | 2.681 | 2.595 | 0.086 | 3.341 | |
| SiC | 08 | 0.30 | 0.75 | −20.54 | 9.463 | 8.439 | 10.857 | 11.123 | 1.361 | 2.684 | 2.595 | 0.089 | 3.434 | |
| SiC | 06 | 0.30 | 1.00 | −20.54 | 9.503 | 8.356 | 10.859 | 11.166 | 1.356 | 2.810 | 2.595 | 0.215 | 8.270 |
Table 5. -point convergence of bulk systems. VBM: valence band maximum; CBM: conduction band minimum. Reference values are the electronic gaps from Table 1. Gap error is .
| System | (Eh) | (mEh) | HOMO PBE | HOMO olLOSC | LUMO PBE | LUMO olLOSC | Gap PBE | Gap olLOSC | Gap Ref. | olLOSC gap error eV | olLOSC gap error % |
| Acridine | -555.067 | -0.935 | -5.426 | -7.809 | -2.976 | -0.830 | 2.449 | 6.979 | 7.394 | -0.415 | -5.618 |
| Anthracene | -539.021 | 0.266 | -4.979 | -7.378 | -2.728 | -0.662 | 2.251 | 6.716 | 7.214 | -0.498 | -6.907 |
| Azulene | -385.478 | 2.300 | -4.970 | -7.582 | -2.881 | -0.432 | 2.089 | 7.149 | 7.003 | 0.146 | 2.089 |
| Benzonitrile | -324.201 | -0.216 | -6.852 | -9.848 | -2.521 | 0.129 | 4.331 | 9.977 | 10.168 | -0.191 | -1.874 |
| Benzoquinone | -381.160 | -0.022 | -6.348 | -10.256 | -4.566 | -1.674 | 1.781 | 8.582 | 8.714 | -0.132 | -1.517 |
| Dichlone | -1453.559 | -1.344 | -6.580 | -9.914 | -4.385 | -2.094 | 2.195 | 7.819 | 8.066 | -0.247 | -3.056 |
| Dinitrobenzonitrile | -733.009 | 0.171 | -7.725 | -11.345 | -4.583 | -1.924 | 3.142 | 9.421 | 9.408 | 0.013 | 0.139 |
| Fumaronitrile | -262.854 | -0.331 | -7.851 | -11.256 | -4.094 | -1.113 | 3.757 | 10.143 | 10.507 | -0.364 | -3.469 |
| Maleicanhydride | -379.039 | 0.071 | -7.129 | -11.342 | -4.199 | -1.083 | 2.930 | 10.259 | 10.308 | -0.049 | -0.477 |
| NDCA | -686.067 | 0.346 | -6.468 | -9.019 | -3.658 | -1.341 | 2.810 | 7.678 | 7.918 | -0.240 | -3.036 |
| Naphthalenedione | -534.682 | 0.036 | -6.151 | -9.668 | -4.171 | -1.676 | 1.979 | 7.992 | 8.403 | -0.411 | -4.888 |
| Nitrobenzene | -436.435 | -0.010 | -6.816 | -10.809 | -3.485 | -0.462 | 3.332 | 10.347 | 9.697 | 0.650 | 6.704 |
| Nitrobenzonitrile | -528.608 | -0.013 | -7.301 | -11.082 | -4.136 | -1.385 | 3.165 | 9.696 | 9.348 | 0.348 | 3.725 |
| Phenazine | -571.106 | 0.090 | -5.735 | -8.998 | -3.404 | -1.110 | 2.331 | 7.887 | 7.387 | 0.500 | 6.773 |
| Phthalicanhydride | -532.560 | -0.428 | -6.900 | -10.598 | -3.601 | -0.945 | 3.299 | 9.653 | 9.700 | -0.047 | -0.485 |
| Phthalimide | -512.703 | -0.751 | -6.350 | -9.990 | -3.299 | -0.674 | 3.051 | 9.315 | 9.482 | -0.167 | -1.757 |
| TCNE | -447.173 | 1.231 | -8.597 | -11.816 | -5.780 | -3.165 | 2.817 | 8.651 | 8.954 | -0.303 | -3.383 |
| TCNQ | -678.022 | 11.383 | -7.050 | -9.031 | -5.514 | -3.587 | 1.536 | 5.444 | 6.262 | -0.818 | -13.070 |
| Tetrachloro-isobenzofuranedione | -2370.300 | 1.541 | -7.106 | -9.671 | -4.062 | -1.763 | 3.044 | 7.908 | 8.430 | -0.522 | -6.191 |
| Tetrachlorobenzoquinone | -2218.912 | 0.299 | -6.983 | -9.490 | -5.010 | -2.631 | 1.973 | 6.859 | 7.762 | -0.903 | -11.637 |
| Tetrafluorobenzenedicarbonitrile | -813.119 | 0.201 | -7.444 | -9.995 | -4.133 | -1.652 | 3.311 | 8.343 | 9.151 | -0.808 | -8.829 |
| Tetrafluorobenzoquinone | -777.913 | 0.071 | -7.374 | -10.361 | -5.127 | -2.348 | 2.248 | 8.013 | 8.861 | -0.848 | -9.571 |
| mDCBN | -416.374 | -0.150 | -7.393 | -10.172 | -3.288 | -0.764 | 4.105 | 9.409 | 9.832 | -0.423 | -4.307 |
Table 8. Large organic molecules olLOSC calculation results with and . (See b and e in Figure 3 and Figure 4) Reference gap (Ref.) is CCSD-T benchmark from Ref[43].
| System | Total energy () | (m) | HOMO (eV) PBE | HOMO (eV) olLOSC | LUMO (eV) PBE | LUMO (eV) olLOSC | Gap (eV) PBE | Gap (eV) olLOSC | Gap (eV) Ref. | olLOSC gap error (eV) | olLOSC gap error (%) |
| CF2 | -237.57 | 0.03 | -7.36 | -12.22 | -3.73 | 0.68 | 3.63 | 12.90 | 12.68 | 0.22 | 1.71 |
| CH3O | -114.96 | 0.05 | -6.06 | -11.18 | -5.34 | -1.01 | 0.73 | 10.17 | 9.51 | 0.66 | 6.93 |
| CN | -92.64 | 1.12 | -9.38 | -14.47 | -7.67 | -3.64 | 1.71 | 10.83 | 10.29 | 0.54 | 5.28 |
| Cl2 | -920.04 | -0.00 | -7.42 | -11.73 | -4.66 | -0.92 | 2.76 | 10.81 | 10.61 | 0.20 | 1.89 |
| H2C | -39.11 | 0.01 | -5.73 | -10.93 | -3.32 | -0.46 | 2.41 | 10.47 | 9.57 | 0.90 | 9.40 |
| H2CS | -437.27 | 0.02 | -5.54 | -9.41 | -3.64 | -0.11 | 1.90 | 9.29 | 9.16 | 0.13 | 1.44 |
| H2N | -55.83 | 0.03 | -7.27 | -12.90 | -4.54 | -0.28 | 2.73 | 12.62 | 11.37 | 1.25 | 10.96 |
| H2P | -342.35 | 0.03 | -6.05 | -10.11 | -4.20 | -1.20 | 1.85 | 8.91 | 8.56 | 0.35 | 4.09 |
| H2Si | -290.46 | 0.07 | -5.86 | -9.43 | -4.06 | -1.24 | 1.80 | 8.19 | 8.48 | -0.29 | -3.42 |
| H3C | -39.79 | 0.00 | -5.42 | -10.44 | -2.95 | 0.24 | 2.48 | 10.68 | 9.84 | 0.84 | 8.52 |
| H3Si | -291.08 | 0.13 | -5.36 | -9.12 | -3.55 | -1.12 | 1.82 | 7.99 | 7.93 | 0.06 | 0.79 |
| HCO | -113.78 | 0.14 | -5.00 | -9.66 | -3.30 | 0.49 | 1.70 | 10.15 | 9.44 | 0.71 | 7.53 |
| HN | -55.17 | 0.03 | -7.91 | -14.23 | -4.32 | 0.02 | 3.60 | 14.26 | 13.15 | 1.11 | 8.40 |
| HO | -75.68 | 0.01 | -7.36 | -14.01 | -6.44 | -1.14 | 0.92 | 12.87 | 11.24 | 1.63 | 14.51 |
| HP | -341.73 | 0.03 | -6.19 | -10.30 | -3.97 | -1.05 | 2.21 | 9.25 | 9.16 | 0.09 | 0.96 |
| HS | -398.59 | 0.02 | -6.22 | -10.53 | -5.79 | -2.24 | 0.43 | 8.29 | 8.05 | 0.24 | 3.01 |
| S2 | -796.07 | -0.01 | -5.91 | -9.56 | -4.61 | -1.30 | 1.30 | 8.26 | 7.92 | 0.34 | 4.29 |
Table 9. Small molecules olLOSC calculation results with and . (See a and d in Figure 3 and Figure 4) Reference gap (Ref.) is CCSD-T benchmark from [43].
| System | Total energy () | () | HOMO (eV) PBE | HOMO (eV) olLOSC | LUMO (eV) PBE | LUMO (eV) olLOSC | Gap (eV) PBE | Gap (eV) olLOSC | Gap (eV) Ref. | olLOSC gap error (eV) | olLOSC gap error (%) |
| CF2 | -237.57 | 0.00 | -7.36 | -12.30 | -3.73 | 0.80 | 3.63 | 13.11 | 12.68 | 0.43 | 3.37 |
| CH3O | -114.96 | 0.01 | -6.06 | -11.26 | -5.34 | -0.90 | 0.73 | 10.35 | 9.51 | 0.84 | 8.84 |
| CN | -92.64 | 0.12 | -9.38 | -14.61 | -7.67 | -3.49 | 1.71 | 11.12 | 10.29 | 0.83 | 8.05 |
| Cl2 | -920.04 | -0.00 | -7.42 | -11.41 | -4.66 | -0.83 | 2.76 | 10.59 | 10.61 | -0.02 | -0.20 |
| H2C | -39.11 | 0.00 | -5.73 | -11.06 | -3.32 | -0.32 | 2.41 | 10.74 | 9.57 | 1.17 | 12.22 |
| H2CS | -437.27 | 0.00 | -5.54 | -9.49 | -3.64 | -0.02 | 1.90 | 9.46 | 9.16 | 0.30 | 3.32 |
| H2N | -55.83 | 0.01 | -7.27 | -13.05 | -4.54 | -0.13 | 2.73 | 12.92 | 11.37 | 1.55 | 13.64 |
| H2P | -342.35 | 0.01 | -6.05 | -10.21 | -4.20 | -1.10 | 1.85 | 9.10 | 8.56 | 0.54 | 6.36 |
| H2Si | -290.46 | 0.02 | -5.86 | -9.51 | -4.06 | -1.16 | 1.80 | 8.34 | 8.48 | -0.14 | -1.63 |
| H3C | -39.79 | 0.00 | -5.42 | -10.57 | -2.95 | 0.36 | 2.48 | 10.93 | 9.84 | 1.09 | 11.05 |
| H3Si | -291.08 | 0.03 | -5.36 | -9.19 | -3.55 | -1.01 | 1.82 | 8.18 | 7.93 | 0.25 | 3.19 |
| HCO | -113.78 | 0.04 | -5.00 | -9.78 | -3.30 | 0.64 | 1.70 | 10.42 | 9.44 | 0.98 | 10.36 |
| HN | -55.17 | 0.01 | -7.91 | -14.41 | -4.32 | 0.17 | 3.60 | 14.59 | 13.15 | 1.44 | 10.91 |
| HO | -75.68 | 0.00 | -7.36 | -14.21 | -6.44 | -0.95 | 0.92 | 13.26 | 11.24 | 2.02 | 17.97 |
| HP | -341.73 | 0.01 | -6.19 | -10.40 | -3.97 | -0.95 | 2.21 | 9.45 | 9.16 | 0.29 | 3.16 |
| HS | -398.59 | 0.00 | -6.22 | -10.65 | -5.79 | -2.13 | 0.43 | 8.53 | 8.05 | 0.48 | 5.94 |
| S2 | -796.07 | -0.00 | -5.91 | -9.64 | -4.61 | -1.22 | 1.30 | 8.43 | 7.92 | 0.51 | 6.39 |
Table 10. Small molecules olLOSC calculation results with and . (See a and d in Figure 3 and Figure 4) Reference gap (Ref.) is CCSD-T benchmark from Ref[43].
| H(C₂H₂)ₙH | LOSC γ | vW | λ | () | ΔE (m) | PBE | olLOSC | Ref. | eV | % |
| Total energy | HOMO (eV) | olLOSC error | ||||||||
| 1 | 0.300 | 0.75 | -78.501 | 0.000 | -6.722 | -10.870 | -10.481 | -0.389 | 3.712 | |
| 1 | 0.477 | 1.00 | -78.501 | 0.000 | -6.722 | -10.975 | -10.481 | -0.494 | 4.712 | |
| 2 | 0.300 | 0.75 | -155.831 | 0.165 | -5.819 | -9.306 | -9.176 | -0.130 | 1.415 | |
| 2 | 0.477 | 1.00 | -155.831 | 0.038 | -5.819 | -9.385 | -9.176 | -0.209 | 2.282 | |
| 3 | 0.300 | 0.75 | -233.162 | 1.026 | -5.366 | -8.390 | -8.177 | -0.213 | 2.599 | |
| 3 | 0.477 | 1.00 | -233.162 | 0.498 | -5.366 | -8.415 | -8.177 | -0.238 | 2.915 | |
| 4 | 0.300 | 0.75 | -310.494 | 2.130 | -5.094 | -7.832 | -7.691 | -0.141 | 1.835 | |
| 4 | 0.477 | 1.00 | -310.494 | 1.173 | -5.094 | -7.877 | -7.691 | -0.186 | 2.414 | |
| 5 | 0.300 | 0.75 | -387.827 | 4.028 | -4.913 | -7.372 | -7.329 | -0.043 | 0.581 | |
| 5 | 0.477 | 1.00 | -387.827 | 2.149 | -4.913 | -7.314 | -7.329 | 0.015 | -0.202 | |
| 6 | 0.300 | 0.75 | -465.160 | 4.596 | -4.783 | -7.076 | -7.044 | -0.032 | 0.451 | |
| 6 | 0.477 | 1.00 | -465.160 | 2.587 | -4.783 | -7.028 | -7.044 | 0.016 | -0.232 | |
| 7 | 0.300 | 0.75 | -542.493 | 6.656 | -4.685 | -6.795 | -6.852 | 0.057 | -0.836 | |
| 7 | 0.477 | 1.00 | -542.493 | 3.667 | -4.685 | -6.764 | -6.852 | 0.088 | -1.285 | |
| 8 | 0.300 | 0.75 | -619.826 | 7.781 | -4.609 | -6.631 | -6.662 | 0.031 | -0.471 | |
| 8 | 0.477 | 1.00 | -619.826 | 4.656 | -4.609 | -6.585 | -6.662 | 0.077 | -1.158 | |
| 9 | 0.300 | 0.75 | -697.159 | 9.879 | -4.548 | -6.440 | -6.555 | 0.115 | -1.751 | |
| 9 | 0.477 | 1.00 | -697.159 | 5.750 | -4.548 | -6.400 | -6.555 | 0.155 | -2.362 | |
| 10 | 0.300 | 0.75 | -774.493 | 10.759 | -4.499 | -6.313 | -6.410 | 0.097 | -1.515 | |
| 10 | 0.477 | 1.00 | -774.493 | 6.569 | -4.499 | -6.263 | -6.410 | 0.147 | -2.294 |
Table 11. All trans-polyacetylene (H(C₂H₂)ₙH olLOSC HOMO energies with { and } and { and } (See a, b, e, f in Figure 7) Reference energies is RASPT2 calculated negative IP energies from Ref[44].

Figure 7. olLOSC errors in ionization energy for all-trans polyacetylene with different LOSC and vW . (a,b): Percentage error for olLOSC within RPA; (c,d): Percentage error for olLOSC beyond RPA; (e,f): Unscaled error for olLOSC within RPA; (g,h): Unscaled error for olLOSC beyond RPA.
| Molecule | Electrons | Basis functions | olLOSC wall time (s) |
| 16 | 130 | 15.15 | |
| 30 | 230 | 34.17 | |
| 44 | 330 | 65.85 | |
| 58 | 430 | 116.92 | |
| 72 | 530 | 188.40 | |
| 86 | 630 | 291.36 | |
| 100 | 730 | 436.64 | |
| 114 | 830 | 638.80 | |
| 128 | 930 | 868.04 | |
| 142 | 1030 | 1209.56 |
Table 14. Wall times in seconds for the olLOSC evaluation for polyacetylene using the cc-pVTZ basis set and cc-pVTZ-RIFIT as the auxiliary basis. The olLOSC method is implemented in QMD and executed on a 32-core Intel(R) Xeon(R) Gold 6226R CPU @ 2.90 GHz, with a NVIDIA A100 GPU employed to accelerate the numerical integration. From the bolded wall times, numerical fitting indicates that the computational cost scales as with or, alternatively, as with .
References
- [1]F. Bloch, “Über die Quantenmechanik der Elektronen in Kristallgittern”, Z. Physik 52, 555–600 (1929).DOI
- [2]G. H. Wannier, “The Structure of Electronic Excitation Levels in Insulating Crystals”, Phys. Rev. 52, 191–197 (1937).DOI
- [3]A. Mahler, J. Z. Williams, N. Q. Su, and W. Yang, Wannier Functions Dually Localized in Space and Energy, Jan. 2022.DOI
- [4]N. Marzari and D. Vanderbilt, “Maximally localized generalized Wannier functions for composite energy bands”, Phys. Rev. B 56, 12847–12865 (1997).DOI
- [5]J. M. Foster and S. F. Boys, “Canonical Configurational Interaction Procedure”, Rev. Mod. Phys. 32, 300–302 (1960).
- [6]S. Baroni, S. de Gironcoli, A. Dal Corso, and P. Giannozzi, “Phonons and related crystal properties from density-functional perturbation theory”, Rev. Mod. Phys. 73, 515–562 (2001).DOI
- [7]I. Timrov, N. Marzari, and M. Cococcioni, “Hubbard parameters from density-functional perturbation theory”, Phys. Rev. B 98, 085127 (2018).DOI
- [8]H. J. Monkhorst and J. D. Pack, “Special points for Brillouin-zone integrations”, Phys. Rev. B 13, 5188–5192 (1976).DOI
- [9]N. Colonna, R. De Gennaro, E. Linscott, and N. Marzari, “Koopmans Spectral Functionals in Periodic Boundary Conditions”, J. Chem. Theory Comput. 18, 5435–5448 (2022).arxiv.org/abs/2202.08155
- [10]R. M. Sternheimer, “Electronic Polarizabilities of Ions from the Hartree-Fock Wave Functions”, Phys. Rev. 96, 951–968 (1954).DOI
- [11]J. Z. Williams and W. Yang, Correcting Delocalization Error in Materials with Localized Orbitals and Linear-Response Screening, July 2024.DOI
- [12]G. L. Oliver and J. P. Perdew, “Spin-density gradient expansion for the kinetic energy”, Phys. Rev. A 20, 397–403 (1979).
- [13]P. K. Chattaraj, D. R. Roy, P. Geerlings, and M. Torrent-Sucarrat, “Local hardness: a critical account”, Theor Chem Account 118, 923–930 (2007).DOI
- [14]F. Della Sala, “Orbital-free methods for plasmonics: Linear response”, J. Chem. Phys. 157, 104101 (2022).DOI
- [15]O. Vahtras, J. Almlöf, and M. W. Feyereisen, “Integral approximations for LCAO-SCF calculations”, Chem. Phys. Lett. 213, 514–518 (1993).DOI
- [16]X. Ren, P. Rinke, V. Blum, J. Wieferink, A. Tkatchenko, A. Sanfilippo, K. Reuter, and M. Scheffler, “Resolution-of-identity approach to Hartree–Fock, hybrid density functionals, RPA, MP2 and GW with numeric atom-centered orbital basis functions”, New J. Phys. 14, 053020 (2012).DOI
- [17]D. M. York and W. Yang, “A chemical potential equalization method for molecular simulations”, J. Chem. Phys. 104, 159–172 (1996).DOI
- [18]N. I. M. Gould, M. E. Hribar, and J. Nocedal, “On the Solution of Equality Constrained Quadratic Programming Problems Arising in Optimization”, SIAM J. Sci. Comput. 23, 1376–1395 (2001).DOI
- [19]S. Hirata and M. Head-Gordon, “Time-dependent density functional theory for radicals An improved description of excited states with substantial double excitation character”, Chem. Phys. Lett. 302, 375–382 (1999).
- [20]W. Wirtinger, “Zur formalen Theorie der Funktionen von mehr komplexen Veränderlichen”, Math. Ann. 97, 357–375 (1927).DOI
- [21]Y. Mei, Z. Chen, and W. Yang, “Self-Consistent Calculation of the Localized Orbital Scaling Correction for Correct Electron Densities and Energy-Level Alignments in Density Functional Theory”, J. Phys. Chem. Lett. 11, 10269–10277 (2020).DOI
- [22]C. Li, X. Zheng, N. Q. Su, and W. Yang, “Localized orbital scaling correction for systematic elimination of delocalization error in density functional approximations”, Nat. Sci. Rev. 5, 203–215 (2018).DOI
- [23]N. Q. Su, A. Mahler, and W. Yang, “Preserving Symmetry and Degeneracy in the Localized Orbital Scaling Correction Approach”, J. Phys. Chem. Lett. 11, 1528–1535 (2020).
- [24]A. Mahler, J. Williams, N. Q. Su, and W. Yang, “Localized orbital scaling correction for periodic systems”, Phys. Rev. B 106, 035147 (2022).DOI
- [25]Y. Fan, J. Z. Williams, and W. Yang, Data and scripts from: ollosc: unified and efficient density functional approximation to correct delocalization error in molecules and periodic materials [data set], Duke Research Data Repository, 2026.DOI
- [26]J. Zheng, Y. Zhao, and D. G. Truhlar, “The dbh24/08 database and its use to assess electronic structure model chemistries for chemical reaction barrier heights”, Journal of Chemical Theory and Computation 5, 808–821 (2009).
- [27]R. Peverati and D. G. Truhlar, “Quest for a universal density functional: the accuracy of density functionals across a broad spectrum of databases in chemistry and physics”, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 372, 10.1098/RSTA.2012.0476 (2014).
- [28]F. Gygi, J.-L. Fattebert, and E. Schwegler, “Computation of Maximally Localized Wannier Functions using a simultaneous diagonalization algorithm”, Computer Physics Communications 155, 1–6 (2003).DOI
- [29]I. Souza, N. Marzari, and D. Vanderbilt, “Maximally localized Wannier functions for entangled energy bands”, Phys. Rev. B 65, 035109 (2001).DOI
- [30]D. Monaco, G. Panati, A. Pisante, and S. Teufel, “Optimal Decay of Wannier functions in Chern and Quantum Hall Insulators”, Commun. Math. Phys. 359, 61–100 (2018).DOI
- [31]V. Vitale, G. Pizzi, A. Marrazzo, J. R. Yates, N. Marzari, and A. A. Mostofi, “Automated high-throughput Wannierization”, npj Comput Mater 6, 1–18 (2020).
- [32]A. Damle, L. Lin, and L. Ying, “Compressed Representation of Kohn–Sham Orbitals via Selected Columns of the Density Matrix”, J. Chem. Theory Comput. 11, 1463–1469 (2015).DOI
- [33]A. Damle, L. Lin, and L. Ying, “SCDM-k: Localized orbitals for solids via selected columns of the density matrix”, Journal of Computational Physics 334, 1–15 (2017).arxiv.org/abs/1507.03354
- [34]E. Prodan and W. Kohn, “Nearsightedness of electronic matter”, PNAS 102, 11635–11638 (2005).pubmed.ncbi.nlm.nih.gov/16087868
- [35]J. Heyd, J. E. Peralta, G. E. Scuseria, and R. L. Martin, “Energy band gaps and lattice parameters evaluated with the Heyd-Scuseria-Ernzerhof screened hybrid functional”, J. Chem. Phys. 123, 174101 (2005).pubmed.ncbi.nlm.nih.gov/16375511
- [36]R. W. G. Wyckoff, Crystal Structures, 2nd (Wiley, New York, 1973).
- [37]F. Tran and P. Blaha, “Accurate Band Gaps of Semiconductors and Insulators with a Semilocal Exchange-Correlation Potential”, Phys. Rev. Lett. 102, 226401 (2009).pubmed.ncbi.nlm.nih.gov/19658882
- [38]R. T. Poole, J. G. Jenkin, J. Liesegang, and R. C. G. Leckey, “Electronic band structure of the alkali halides. I. Experimental parameters”, Phys. Rev. B 11, 5179–5189 (1975).
- [39]O. Madelung, Semiconductors: Data Handbook (Springer Berlin Heidelberg, Berlin, Heidelberg, 2004).DOI
- [40]M. Engel, H. Miranda, L. Chaput, A. Togo, C. Verdi, M. Marsman, and G. Kresse, “Zero-point renormalization of the band gap of semiconductors and insulators using the projector augmented wave method”, Phys. Rev. B 106, 094316 (2022).DOI
- [41]A. Miglio, V. Brousseau-Couture, E. Godbout, G. Antonius, Y.-H. Chan, S. G. Louie, M. Côté, M. Giantomassi, and X. Gonze, “Predominance of non-adiabatic effects in zero-point renormalization of the electronic band gap”, npj Comput Mater 6, 1–8 (2020).DOI
- [42]H. Shang, J. Zhao, and J. Yang, “Assessment of the Mass Factor for the Electron–Phonon Coupling in Solids”, J. Phys. Chem. C 125, 6479–6485 (2021).DOI
- [43]R. M. Richard, M. S. Marshall, O. Dolgounitcheva, J. V. Ortiz, J. L. Brédas, N. Marom, and C. D. Sherrill, “Accurate ionization potentials and electron affinities of acceptor molecules i. reference data at the ccsd(t) complete basis set limit”, Journal of Chemical Theory and Computation 12, 595–604 (2016).DOI
- [44]A. R. M. Shahi, C. J. Cramer, and L. Gagliardi, “Second-order perturbation theory with complete and restricted active space reference functions applied to oligomeric unsaturated hydrocarbons”, Physical Chemistry Chemical Physics 11, 10964–10972 (2009).DOI
