Orbital-Free Density Functional Theory: An Attractive Electronic Structure Method for Large-Scale First-Principles Simulations.

Kohn-Sham Density Functional Theory (KSDFT) is the most widely used electronic structure method in chemistry, physics, and materials science, with thousands of calculations cited annually. This ubiquity is rooted in the favorable accuracy vs cost balance of KSDFT. Nonetheless, the ambitions and expectations of researchers for use of KSDFT in predictive simulations of large, complicated molecular systems are confronted with an intrinsic computational cost-scaling challenge. Particularly evident in the context of first-principles molecular dynamics, the challenge is the high cost-scaling associated with the computation of the Kohn-Sham orbitals. Orbital-free DFT (OFDFT), as the name suggests, circumvents entirely the explicit use of those orbitals. Without them, the structural and algorithmic complexity of KSDFT simplifies dramatically and near-linear scaling with system size irrespective of system state is achievable. Thus, much larger system sizes and longer simulation time scales (compared to conventional KSDFT) become accessible; hence, new chemical phenomena and new materials can be explored. In this review, we introduce the historical contexts of OFDFT, its theoretical basis, and the challenge of realizing its promise via approximate kinetic energy density functionals (KEDFs). We review recent progress on that challenge for an array of KEDFs, such as one-point, two-point, and machine-learnt, as well as some less explored forms. We emphasize use of exact constraints and the inevitability of design choices. Then, we survey the associated numerical techniques and implemented algorithms specific to OFDFT. We conclude with an illustrative sample of applications to showcase the power of OFDFT in materials science, chemistry, and physics.

"Atomic Topping" of MnOx on Al2O3 to Create Electron-Rich, Aperiodic, Lattice Oxygens that Resemble Noble Metals for Catalytic Oxidation.

Enhancing the catalytic oxidation activity of traditional transition-metal oxides to rival that of noble metals has been a prominent focus in the field of catalysis. However, existing synthesis strategies that focus on controlling the electronic states of metal centers have not yet fully succeeded in achieving this goal. Our current research reveals that manipulating the electronic states of oxygen centers can yield unexpected results. By creating electron-rich, aperiodic lattice oxygens through atomic topping of MnOx, we have produced a catalyst with performance that closely resembles supported Pt. Spherical aberration-corrected transmission electron microscopy and X-ray absorption spectra have confirmed that the atomic topping of the MnOx layer on Al2O3 can form an aperiodic arrangement oxide structure. Near-ambient pressure X-ray photoelectron spectroscopy, in situ diffuse reflectance infrared Fourier transform spectroscopy, reaction kinetics test, and theoretical calculations demonstrated that this structure significantly increases the electron density around the oxygen in MnOx, shifting the activation center for CO adsorption from Mn to O, thereby exhibiting catalytic activity and stability close to that of the precious metal Pt. This study presents a fresh perspective on designing efficient oxide catalysts by targeting electron-rich anionic centers, thereby deepening the understanding of how these centers can be altered to enhance catalytic efficiency in oxidation reactions.

Entropy is a good approximation to the electronic (static) correlation energy.

For an electronic system, given a mean field method and a distribution of orbital occupation numbers that are close to the natural occupations of the correlated system, we provide formal evidence and computational support to the hypothesis that the entropy (or more precisely -σS, where σ is a parameter and S is the entropy) of such a distribution is a good approximation to the correlation energy. Underpinning the formal evidence are mild assumptions: the correlation energy is strictly a functional of the occupation numbers, and the occupation numbers derive from an invertible distribution. Computational support centers around employing different mean field methods and occupation number distributions (Fermi-Dirac, Gaussian, and linear), for which our claims are verified for a series of pilot calculations involving bond breaking and chemical reactions. This work establishes a formal footing for those methods employing entropy as a measure of electronic correlation energy (e.g., i-DMFT [Wang and Baerends, Phys. Rev. Lett. 128, 013001 (2022)] and TAO-DFT [J.-D. Chai, J. Chem. Phys. 136, 154104 (2012)]) and sets the stage for the widespread use of entropy functionals for approximating the (static) electronic correlation.

Nonadiabatic couplings from a variational excited state method based on constrained DFT.

Excited Costrained Density Functional Theory (XCDFT) [Ramos and Pavanello, J. Chem. Phys. 148, 144103 (2018)] is a variational excited state method that extends ground state DFT to the computation of low-lying excited states. It borrows much of the machinery of Constrained DFT (CDFT) with a crucial difference: the constraint imposes a population of one electron in the Hilbert space spanned by the virtuals of a reference ground state. In this work, we present theory and implementation for evaluating nonadiabatic coupling vectors (NACVs) between the first excited state computed with XCDFT and the ground state. Our NACVs are computed analytically using density functional perturbation theory with a formalism that is general enough that could be applied to CDFT diabatic states. We showcase the new method with pilot NACV calculations for the conical intersection in H3, the avoided crossing in selenoacrolein, and the NACV magnitudes in azobenzene. Despite complications from the nonorthogonality of the wavefunctions, XCDFT's energy surfaces and NACVs reproduce benchmark values and respect known sum rules within a reasonable degree. This shows that XCDFT is a viable method for nonadiabatic dynamics simulations.

Tuning the electronic properties of the γ-Al2O3 surface by phosphorus doping.

Tuning the electronic properties of oxide surfaces is of pivotal importance, because they find applicability in a variety of industrial processes, including catalysis. Currently, the industrial protocols for synthesizing oxide surfaces are limited to only partial control of the oxide's properties. This is because the ceramic processes result in complex morphologies and a priori unpredictable behavior of the products. While the bulk doping of alumina surfaces has been demonstrated to enhance their catalytic applications (i.e. hydrodesulphurization (HDS)), the fundamental understanding of this phenomenon and its effect at an atomic level remain unexplored. In our joint experimental and computational study, simulations based on Density Functional Theory (DFT), synthesis, and a variety of surface characterization techniques are exploited for the specific goal of understanding the structure-function relationship of phosphorus-doped γ-Al2O3 surfaces. Our theoretical calculations and experimental results agree in finding that P doping of γ-Al2O3 leads to a significant decrease in its work function. Our computational models show that this decrease is due to the formation of a new surface dipole, providing a clear picture of the effect of P doping at the surface of γ-Al2O3. In this study, we uncover a general paradigm for tuning support-catalyst interactions that involves electrostatic properties of doped γ-Al2O3 surface, specifically the surface dipole. Our findings open a new pathway for engineering the electronic properties of metal oxides' surfaces.

Low-lying excited states by constrained DFT.

Exploiting the machinery of Constrained Density Functional Theory (CDFT), we propose a variational method for calculating low-lying excited states of molecular systems. We dub this method eXcited CDFT (XCDFT). Excited states are obtained by self-consistently constraining a user-defined population of electrons, Nc, in the virtual space of a reference set of occupied orbitals. By imposing this population to be Nc = 1.0, we computed the first excited state of 15 molecules from a test set. Our results show that XCDFT achieves an accuracy in the predicted excitation energy only slightly worse than linear-response time-dependent DFT (TDDFT), but without incurring into problems of variational collapse typical of the more commonly adopted ΔSCF method. In addition, we selected a few challenging processes to test the limits of applicability of XCDFT. We find that in contrast to TDDFT, XCDFT is capable of reproducing energy surfaces featuring conical intersections (azobenzene and H3) with correct topology and correct overall energetics also away from the intersection. Venturing to condensed-phase systems, XCDFT reproduces the TDDFT solvatochromic shift of benzaldehyde when it is embedded by a cluster of water molecules. Thus, we find XCDFT to be a competitive method among single-reference methods for computations of excited states in terms of time to solution, rate of convergence, and accuracy of the result.

Nonlocal kinetic energy functionals by functional integration.

Since the seminal studies of Thomas and Fermi, researchers in the Density-Functional Theory (DFT) community are searching for accurate electron density functionals. Arguably, the toughest functional to approximate is the noninteracting kinetic energy, Ts[ρ], the subject of this work. The typical paradigm is to first approximate the energy functional and then take its functional derivative, δTs[ρ]δρ(r), yielding a potential that can be used in orbital-free DFT or subsystem DFT simulations. Here, this paradigm is challenged by constructing the potential from the second-functional derivative via functional integration. A new nonlocal functional for Ts[ρ] is prescribed [which we dub Mi-Genova-Pavanello (MGP)] having a density independent kernel. MGP is constructed to satisfy three exact conditions: (1) a nonzero "Kinetic electron" arising from a nonzero exchange hole; (2) the second functional derivative must reduce to the inverse Lindhard function in the limit of homogenous densities; (3) the potential is derived from functional integration of the second functional derivative. Pilot calculations show that MGP is capable of reproducing accurate equilibrium volumes, bulk moduli, total energy, and electron densities for metallic (body-centered cubic, face-centered cubic) and semiconducting (crystal diamond) phases of silicon as well as of III-V semiconductors. The MGP functional is found to be numerically stable typically reaching self-consistency within 12 iterations of a truncated Newton minimization algorithm. MGP's computational cost and memory requirements are low and comparable to the Wang-Teter nonlocal functional or any generalized gradient approximation functional.

Constrained subsystem density functional theory.

Constrained Subsystem Density Functional Theory (CSDFT) allows to compute diabatic states for charge transfer reactions using the machinery of the constrained DFT method, and at the same time is able to embed such diabatic states in a molecular environment via a subsystem DFT scheme. The CSDFT acronym is chosen to reflect the fact that on top of the subsystem DFT approach, a constraining potential is applied to each subsystem. We show that CSDFT can successfully tackle systems as complex as single stranded DNA complete of its backbone, and generate diabatic states as exotic as a hole localized on a phosphate group as well as on the nucleobases. CSDFT will be useful to investigators needing to evaluate the environmental effect on charge transfer couplings for systems in condensed phase environments.

Revealing electronic open quantum systems with subsystem TDDFT.

Open quantum systems (OQSs) are perhaps the most realistic systems one can approach through simulations. In recent years, describing OQSs with Density Functional Theory (DFT) has been a prominent avenue of research with most approaches based on a density matrix partitioning in conjunction with an ad-hoc description of system-bath interactions. We propose a different theoretical approach to OQSs based on partitioning of the electron density. Employing the machinery of subsystem DFT (and its time-dependent extension), we provide a novel way of isolating and analyzing the various terms contributing to the coupling between the system and the surrounding bath. To illustrate the theory, we provide numerical simulations on a toy system (a molecular dimer) and on a condensed phase system (solvated excimer). The simulations show that non-Markovian dynamics in the electronic system-bath interactions are important in chemical applications. For instance, we show that the superexchange mechanism of transport in donor-bridge-acceptor systems is a non-Markovian interaction between the donor-acceptor (OQS) with the bridge (bath) which is fully characterized by real-time subsystem time-dependent DFT.

Avoiding fractional electrons in subsystem DFT based ab-initio molecular dynamics yields accurate models for liquid water and solvated OH radical.

In this work we achieve three milestones: (1) we present a subsystem DFT method capable of running ab-initio molecular dynamics simulations accurately and efficiently. (2) In order to rid the simulations of inter-molecular self-interaction error, we exploit the ability of semilocal frozen density embedding formulation of subsystem DFT to represent the total electron density as a sum of localized subsystem electron densities that are constrained to integrate to a preset, constant number of electrons; the success of the method relies on the fact that employed semilocal nonadditive kinetic energy functionals effectively cancel out errors in semilocal exchange-correlation potentials that are linked to static correlation effects and self-interaction. (3) We demonstrate this concept by simulating liquid water and solvated OH(•) radical. While the bulk of our simulations have been performed on a periodic box containing 64 independent water molecules for 52 ps, we also simulated a box containing 256 water molecules for 22 ps. The results show that, provided one employs an accurate nonadditive kinetic energy functional, the dynamics of liquid water and OH(•) radical are in semiquantitative agreement with experimental results or higher-level electronic structure calculations. Our assessments are based upon comparisons of radial and angular distribution functions as well as the diffusion coefficient of the liquid.

Exact kinetic energy enables accurate evaluation of weak interactions by the FDE-vdW method.

The correlation energy of interaction is an elusive and sought-after interaction between molecular systems. By partitioning the response function of the system into subsystem contributions, the Frozen Density Embedding (FDE)-vdW method provides a computationally amenable nonlocal correlation functional based on the adiabatic connection fluctuation dissipation theorem applied to subsystem density functional theory. In reproducing potential energy surfaces of weakly interacting dimers, we show that FDE-vdW, either employing semilocal or exact nonadditive kinetic energy functionals, is in quantitative agreement with high-accuracy coupled cluster calculations (overall mean unsigned error of 0.5 kcal/mol). When employing the exact kinetic energy (which we term the Kohn-Sham (KS)-vdW method), the binding energies are generally closer to the benchmark, and the energy surfaces are also smoother.

Subsystem real-time time dependent density functional theory.

We present the extension of Frozen Density Embedding (FDE) formulation of subsystem Density Functional Theory (DFT) to real-time Time Dependent Density Functional Theory (rt-TDDFT). FDE is a DFT-in-DFT embedding method that allows to partition a larger Kohn-Sham system into a set of smaller, coupled Kohn-Sham systems. Additional to the computational advantage, FDE provides physical insight into the properties of embedded systems and the coupling interactions between them. The extension to rt-TDDFT is done straightforwardly by evolving the Kohn-Sham subsystems in time simultaneously, while updating the embedding potential between the systems at every time step. Two main applications are presented: the explicit excitation energy transfer in real time between subsystems is demonstrated for the case of the Na4 cluster and the effect of the embedding on optical spectra of coupled chromophores. In particular, the importance of including the full dynamic response in the embedding potential is demonstrated.

Self-assembled organic nanowires for high power density lithium ion batteries.

The electroactive organic materials are promising alternatives to inorganic electrode materials for the new generation of green Li-ion batteries due to their sustainability, environmental benignity, and low cost. Croconic acid disodium salt (CADS) was used as Li-ion battery electrode, and CADS organic wires with different diameters were fabricated through a facile synthetic route using antisolvent crystallization method to overcome the challenges of low electronic conductivity of CADS and lithiation induced strain. The CADS nanowire exhibits much better electrochemical performance than its crystal bulk material and microwire counterpart. CADS nanowire with a diameter of 150 nm delivers a reversible capability of 177 mAh g(-1) at a current density of 0.2 C and retains capacity of 170 mAh g(-1) after 110 charge/discharge cycles. The nanowire structure also remarkably enhances the kinetics of croconic acid disodium salt. The CADS nanowire retains 50% of the 0.1 C capacity even when the current density increases to 6 C. In contrast, the crystal bulk and microwire material completely lose their capacities when the current density merely increases to 2 C. Such a high rate performance of CADS nanowire is attributed to its short ion diffusion pathway and large surface area, which enable fast ion and electron transport in the electrode. The theoretical calculation suggests that lithiation of CADS experiences an ion exchange process. The sodium ions in CADS will be gradually replaced by lithium ions during the lithiation and delithiation of CADS electrode, which is confirmed by inductively coupled plasma test.

Periodic subsystem density-functional theory.

By partitioning the electron density into subsystem contributions, the Frozen Density Embedding (FDE) formulation of subsystem Density Functional Theory (DFT) has recently emerged as a powerful tool for reducing the computational scaling of Kohn-Sham DFT. To date, however, FDE has been employed to molecular systems only. Periodic systems, such as metals, semiconductors, and other crystalline solids have been outside the applicability of FDE, mostly because of the lack of a periodic FDE implementation. To fill this gap, in this work we aim at extending FDE to treat subsystems of molecular and periodic character. This goal is achieved by a dual approach. On one side, the development of a theoretical framework for periodic subsystem DFT. On the other, the realization of the method into a parallel computer code. We find that periodic FDE is capable of reproducing total electron densities and (to a lesser extent) also interaction energies of molecular systems weakly interacting with metallic surfaces. In the pilot calculations considered, we find that FDE fails in those cases where there is appreciable density overlap between the subsystems. Conversely, we find FDE to be in semiquantitative agreement with Kohn-Sham DFT when the inter-subsystem density overlap is low. We also conclude that to make FDE a suitable method for describing molecular adsorption at surfaces, kinetic energy density functionals that go beyond the GGA level must be employed.

FDE-vdW: A van der Waals inclusive subsystem density-functional theory.

We present a formally exact van der Waals inclusive electronic structure theory, called FDE-vdW, based on the Frozen Density Embedding formulation of subsystem Density-Functional Theory. In subsystem DFT, the energy functional is composed of subsystem additive and non-additive terms. We show that an appropriate definition of the long-range correlation energy is given by the value of the non-additive correlation functional. This functional is evaluated using the fluctuation-dissipation theorem aided by a formally exact decomposition of the response functions into subsystem contributions. FDE-vdW is derived in detail and several approximate schemes are proposed, which lead to practical implementations of the method. We show that FDE-vdW is Casimir-Polder consistent, i.e., it reduces to the generalized Casimir-Polder formula for asymptotic inter-subsystems separations. Pilot calculations of binding energies of 13 weakly bound complexes singled out from the S22 set show a dramatic improvement upon semilocal subsystem DFT, provided that an appropriate exchange functional is employed. The convergence of FDE-vdW with basis set size is discussed, as well as its dependence on the choice of associated density functional approximant.

Describing long-range charge-separation processes with subsystem density-functional theory.

Long-range charge-transfer processes in extended systems are difficult to describe with quantum chemical methods. In particular, cost-effective (non-hybrid) approximations within time-dependent density functional theory (DFT) are not applicable unless special precautions are taken. Here, we show that the efficient subsystem DFT can be employed as a constrained DFT variant to describe the energetics of long-range charge-separation processes. A formal analysis of the energy components in subsystem DFT for such excitation energies is presented, which demonstrates that both the distance dependence and the long-range limit are correctly described. In addition, electronic couplings for these processes as needed for rate constants in Marcus theory can be obtained from this method. It is shown that the electronic structure of charge-separated states constructed by a positively charged subsystem interacting with a negatively charged one is difficult to converge - charge leaking from the negative subsystem to the positive one can occur. This problem is related to the delocalization error in DFT and can be overcome with asymptotically correct exchange-correlation (XC) potentials or XC potentials including a sufficiently large amount of exact exchange. We also outline an approximate way to obtain charge-transfer couplings between locally excited and charge-separated states.

Ab initio molecular dynamics calculations on scattering of hyperthermal H atoms from Cu(111) and Au(111).

Energy loss from the translational motion of an atom or molecule impinging on a metal surface to the surface may determine whether the incident particle can trap on the surface, and whether it has enough energy left to react with another molecule present at the surface. Although this is relevant to heterogeneous catalysis, the relative extent to which energy loss of hot atoms takes place to phonons or electron-hole pair (ehp) excitation, and its dependence on the system's parameters, remain largely unknown. We address these questions for two systems that present an extreme case of the mass ratio of the incident atom to the surface atom, i.e., H + Cu(111) and H + Au(111), by presenting adiabatic ab initio molecular dynamics (AIMD) predictions of the energy loss and angular distributions for an incidence energy of 5 eV. The results are compared to the results of AIMDEFp calculations modeling energy loss to ehp excitation using an electronic friction ("EF") model applied to the AIMD trajectories, so that the energy loss to the electrons is calculated "post" ("p") the computation of the AIMD trajectory. The AIMD calculations predict average energy losses of 0.38 eV for Cu(111) and 0.13-0.14 eV for Au(111) for H-atoms that scatter from these surfaces without penetrating the surface. These energies closely correspond with energy losses predicted with Baule models, which is suggestive of structure scattering. The predicted adiabatic integral energy loss spectra (integrated over all final scattering angles) all display a lowest energy peak at an energy corresponding to approximately 80% of the average adiabatic energy loss for non-penetrative scattering. In the adiabatic limit, this suggests a way of determining the approximate average energy loss of non-penetratively scattered H-atoms from the integral energy loss spectrum of all scattered H-atoms. The AIMDEFp calculations predict that in each case the lowest energy loss peak should show additional energy loss in the range 0.2-0.3 eV due to ehp excitation, which should be possible to observe. The average non-adiabatic energy losses for non-penetrative scattering exceed the adiabatic losses to phonons by 0.9-1.0 eV. This suggests that for scattering of hyperthermal H-atoms from coinage metals the dominant energy dissipation channel should be to ehp excitation. These predictions can be tested by experiments that combine techniques for generating H-atom beams that are well resolved in translational energy and for detecting the scattered atoms with high energy-resolution.

Communication: Visible line intensities of the triatomic hydrogen ion from experiment and theory.

The visible spectrum of H3(+) is studied using high-sensitivity action spectroscopy in a cryogenic radiofrequency multipole trap. Advances are made to measure the weak ro-vibrational transitions from the lowest rotational states of H3(+) up to high excitation energies providing visible line intensities and, after normalisation to an infrared calibration line, the corresponding Einstein B coefficients. Ab initio predictions for the Einstein B coefficients are obtained from a highly precise dipole moment surface of H3(+) and found to be in excellent agreement, even in the region where states have been classified as chaotic.

Unraveling the Hydration Shell Structure and Dynamics of Group 10 Aqua Ions.

We present ab initio simulations based on subsystem DFT of group 10 aqua ions accurately compared against experimental data on hydration structure. Our simulations provide insights into the molecular structures and dynamics of hydration shells, offering recalibrated interpretations of experimental results. We observe a soft, but distinct second hydration shell in Palladium (Pd) due to a balance between thermal fluctuations, metal-water interactions, and hydrogen bonding. Nickel (Ni) and platinum (Pt) exhibit more rigid hydration shells. Notably, our simulations align with experimental findings for Pd, showing axial hydration marked by a broad peak at about 3 Å in the Pd-O radial distribution function, revising the previously sharp "mesoshell" prediction. We introduce the "hydrogen bond dome" concept to describe a resilient network of hydrogen-bonded water molecules around the metal, which plays a critical role in the axial hydration dynamics.