The aspiration of abinitio quantum chemistry—to predict molecular properties from nothing more than fundamental physical constants and the atomic composition of a system—represents one of the most ambitious programs in computational science. The Latin phrase “from the beginning” carries a promise: that armed only with the masses of electrons and nuclei, the magnitude of the elementary charge, and Planck’s constant, etc., one can, in principle, predict the color of a transition metal complex, the free energy of a biochemical reaction, or the excited-state dynamics of a photocatalyst. This promise of generalizability and accuracy distinguishes abinitio methods from semi-empirical approaches or molecular mechanics force fields, which rely on parameters fitted to experimental data. Yet the reality of abinitio quantum chemistry is more nuanced than this idealized vision suggests (see Figure 1Figure 1).

The central focus of this review is that modern abinitio methods are not constructed from a single, monolithic physical theory, but rather emerge from a carefully orchestrated synthesis of distinct foundational theories: classical mechanics, quantum mechanics, electromagnetism, thermodynamics and statistical mechanics, relativity, and quantum field theory. Each theory contributes essential conceptual frameworks and mathematical formalisms that enable practical calculation, yet each simultaneously introduces approximations, limitations, or constraints that define the boundaries of predictive accuracy. The Born-Oppenheimer approximation, for instance, is indispensable for reducing computational complexity, yet it originates from a classical mechanical picture that fails for non-adiabatic processes [1][2]. The Coulombic potential terms that govern molecular stability are expressions of classical electromagnetism, yet they treat the electromagnetic field as a static background rather than a dynamical quantum entity [3]. Even the gold-standard Coupled Cluster methods, celebrated for their systematic improvability, rest upon the formal apparatus of non-relativistic quantum field theory while pragmatically truncating the many-body expansion [4].

Understanding this interdependent hierarchy of physical theories is not merely an exercise in historical or philosophical taxonomy. Rather, it provides critical insight into why certain approximations work, when they fail, and how computational chemistry must evolve to extend its predictive reach. The field stands at a pivotal juncture: as chemists probe heavier elements [5]-[7], investigate molecules in optical cavities [8], or demand thermodynamic accuracy approaching chemical precision [9], the approximations that once rendered calculations tractable—the neglect of relativity, the classical treatment of nuclear motion, the static electromagnetic field—become the dominant sources of error. The path forward requires systematically replacing these convenience-driven approximations with more rigorous, unified frameworks, even at significant computational expense.

Figure 1.Ab initio molecular simulation methods summary.

We begin with Classical Mechanics, examining how it enables the Born-Oppenheimer approximation and molecular dynamics simulations, while constraining our ability to describe coupled electronic-nuclear phenomena [1][2][10]. Next, we detail how quantum mechanics and classical electromagnetism together define the molecular Hamiltonian, and how the many-body problem this creates necessitates sophisticated approximations [1][11]. We then explore the role of thermodynamics and statistical mechanics in bridging microscopic quantum states to macroscopic observables through partition functions, free energy methods, and ensemble sampling [9][12]. Subsequently, we examine the mandatory inclusion of relativity for heavy elements, which fundamentally alters the kinetic energy operator and introduces spin-dependent effects [5]-[7]. Finally, we discuss how Quantum Field Theory provides both the formal language (second quantization) underpinning high-accuracy wave function methods and the framework for the emerging frontier of quantum electrodynamics in chemistry, where the electromagnetic field itself is quantized [4][8][13].

Throughout, we emphasize the central trade-off governing all abinitio work: the tension between rigor and computational feasibility. Methods derived from more complete physical theories invariably demand greater computational resources, scaling steeply with system size. The ongoing evolution of the field can thus be viewed as a systematic campaign to incorporate increasingly fundamental physics—from non-relativistic to relativistic descriptions [5][6], from mean-field to correlated treatments [3][4], from classical to quantized fields [8][13]—while developing the mathematical and computational tools necessary to make such calculations tractable. The result is not a single “abinitio method” but a spectrum of approaches, each representing a particular compromise between physical completeness and practical utility, each appropriate for different chemical questions and accuracy requirements.

Classical Mechanics, formulated in the seventeenth century, successfully describes the motion of macroscopic objects and bulk matter, such as projectiles, planets, and machinery. However, it was definitively superseded by quantum mechanics for microscopic systems, as it fails to account for particle-wave duality, the uncertainty principle, or the quantization and energy-matter relationship exhibited by particles like electrons and atoms. CM is now understood as an approximate theory to the more generalized framework of Quantum Mechanics, valid only for non-relativistic, low-energy particles in weak gravitational fields. Despite its inadequacy at the quantum level, CM remains central to making abinitio calculations tractable, primarily through its influence on how the massive atomic nuclei and their motions are treated within a molecular system [1][3][12].

The governing equation of classical dynamics is Newton’s second law. For a system of N nuclei, the equation of motion for the I-th nucleus with mass M I and position R I is determined by F = M I R ¨ I = − ∇ I E e l ( R ) , where R ¨ I is the acceleration of the nucleus, and − ∇ I E e l ( R ) is the gradient of potential acting on it. Numerically integrating this equation over time generates a trajectory of nuclear positions, a technique known as Ab Initio Molecular Dynamics (AIMD) [10]. This approach provides a powerful link to macroscopic observables; by propagating these classical equations of motion and applying the principles of statistical mechanics, one can compute thermodynamic properties such as free energies and reaction rates from first principles [9][12].

Another most significant contribution of CM to computational chemistry is the Born-Oppenheimer (BO) approximation [3]. This approximation is physically justified by the enormous difference in mass between the nuclei and the electrons ( M nucleus ≫ m electron ). Because the nuclei move much more slowly than the light electrons, the electronic wavefunction is assumed to adjust instantaneously to the fixed nuclear positions. Mathematically, the BO approximation allows the total molecular Hamiltonian to be separated into independent electronic and nuclear components. By treating the nuclei as fixed classical coordinates during the electronic structure calculation, the nuclear kinetic energy term ( T N ) is omitted. This simplification drastically reduces the dimensionality and complexity of the electronic Schrödinger equation, making the problem more solvable [1][4]. The resulting eigenvalue (usually energy) obtained from the electronic structure calculation, which includes kinetic and potential contributions from electrons and fixed nuclei, serves as the Potential Energy Surface (PES) that governs the subsequent motion of the nuclei. Once determined within the BO framework, this PES defines the energy landscape for nuclear degrees of freedom, which are then often simulated using classical or semi-classical methods such as molecular dynamics [2][10]. However, the BO approximation, while foundational, introduces a fundamental constraint rooted in classical mechanics: the assumption of adiabaticity, meaning the electronic state remains unchanged as nuclei move. This CM-inspired constraint is the primary source of error in describing phenomena where electronic and nuclear motions are strongly coupled (non-adiabatic effects), such as in charge transfer or dynamics near conical intersections [2]. One illustrative chemical example where this approximation breaks down is the non-adiabatic quenching of electronically excited hydroxyl radicals by hydrogen. In this reaction, transitions between multiple electronic states occur during the collision, mediated by strong non-adiabatic couplings near conical intersections where electronic and nuclear motions cannot be treated independently [14]. Consequently, the success of standard abinitio methods is conditional upon the chemical system being adequately described by this classical separation of nuclear and electronic motion.

Marx and Hutter (2000) demonstrated the power of Ab Initio Molecular Dynamics (AIMD) by performing extensive Car-Parrinello MD simulations of liquid water at ambient conditions [15]. By propagating nuclear trajectories according to classical Newtonian dynamics while computing forces from Density Functional Theory (DFT) electronic structure calculations at each time step, they obtained radial distribution functions, diffusion coefficients, and hydrogen bond dynamics in excellent agreement with experiments.

Liang and Lipscomb (1988) employed a hybrid Quantum Mechanics/Molecular Mechanics (QM/MM) approach called PRDDO (Partial Retention of Diatomic Differential Overlap) combined with classical molecular dynamics to elucidate the complete catalytic cycle of carbonic anhydrase, one of nature’s most efficient enzymes [16]. The active site was treated quantum mechanically using abinitio methods, while the protein environment and solvent were described by classical force fields. Classical Newtonian equations of motion governed all nuclear coordinates, enabling the generation of extensive conformational sampling necessary for free energy calculations. Using sampling and thermodynamic integration along reaction coordinates for protein dynamics, they computed activation free energies for each elementary step, identifying a rate-limiting proton transfer with an energy barrier of 13.2 kcal/mol—in great agreement with the experimental value of around 10 kcal/mol derived from kinetic measurements.

Markland and Ceriotti (2018) systematically investigated the failure of classical treatments by comparing classical MD with Path Integral Molecular Dynamics (PIMD), which incorporates nuclear quantum effects [17]. In biological systems, PIMD simulations including Nuclear Quantum Effects cause a 10,000-fold change in the acidity constant of a key tyrosine active-site residue. Similarly, the isotope effect between ⁷LiH and ⁷LiD was studied from classical simulations versus quantum treatment that accounts for quantum effects on nuclear motion in Molecular Dynamics (MD) simulations, and the experimental isotopic shift in the lattice parameter is reproduced, which supports the latter value. In addition, for H₂ and D₂ diffusion in materials, classical dynamics usually underestimate or overestimate diffusion rate under different temperature ranges, reflecting the failure to capture quantum nuclear effects from nuclei. These findings establish qualitative and quantitative criteria for the validity of classical nuclear approximations.

Thermodynamics and Statistical Mechanics (SM) provide the essential conceptual framework for relating the microscopic, quantum mechanical properties of matter to the macroscopic, bulk properties observed in chemical experiments [12]. SM operates on the premise that macroscopic properties like internal energy (E), enthalpy (H), entropy (S), and Gibbs free energy (G = H − TS) are determined by the statistical distribution of a system’s vast number of microscopic energy states at a given temperature [12]. The fundamental link between the microscopic Hamiltonian operator (H) and the bulk thermodynamic properties is the Canonical Partition Function (Q). For a discrete quantum mechanical system, Q is defined as the trace of the Boltzmann factor, Q = T r ( e − β E ) = ∑ i g i e − β E i , where β = 1 / k B T [12]. High-accuracy abinitio calculations are used to determine the necessary energy levels and potentials, which are then integrated into SM formalisms (often via path integral methods) to calculate macroscopic thermodynamic properties.

Abinitio calculations of Gibbs energy involve two major steps. First, the electronic energy is computed using a method like Hartree-Fock (HF), Density Functional Theory (DFT), or Coupled Cluster (CC) [1][4]. Second, statistical mechanical techniques are applied to compute thermal corrections for the translational, rotational, and vibrational degrees of freedom [12]. These corrections require statistical sampling or integration over ensemble properties. The principles of TD are also leveraged in advanced computational techniques, notably alchemical free energy calculations [9]. These methods exploit the fact that free energy is a state function, independent of the path taken between two states [9]. By constructing “bridging” potential energy functions that represent non-physical, alchemical intermediate states, these methods compute the free energy difference associated with complex transfer processes. This computational strategy allows for the efficient determination of transfer free energies with orders of magnitude less simulation time than is required for direct molecular dynamics simulations of the physical process. The accurate calculation of free energy requires not only precise potential energy calculation but also adequate sampling of the configurational space to capture the statistical mechanical ensemble properties for macroscopic effects. Monte Carlo (MC) and Molecular Dynamics (MD) are primarily applied for configuration generation, and both methods serve as the foundation for the free energy perturbation and thermodynamic integration techniques, which bridge the microscopic energy into macroscopic effects [9][12].

Monte Carlo methods generate configurations of a molecular system by making random moves that are accepted or rejected based on statistical criteria. The most widely used algorithm is the Metropolis-Hastings algorithm, which ensures that the generated configurations follow the correct statistical distribution. The core principle of MC sampling is based on importance sampling, where configurations are generated with a probability proportional to their Boltzmann weight: P ( r N ) ∝ exp ( − β U ( r N ) ) , where β = 1 K B T , U ( r N ) is the potential energy of configuration r N , and K B is the Boltzmann constant. The Metropolis acceptance criterion determines whether a proposed move from state i to state j is accepted: P a c c ( i → j ) = min ( 1 , exp [ − β ( U ( r j N ) − U ( r i N ) ) ] ) [12][18].

Molecular Dynamics simulations solve Newton’s equations of motion for all atoms in the system, generating trajectories that sample phase space according to the microcanonical (NVE), canonical (NVT), or isothermal-isobaric (NPT) ensemble when appropriate thermostats and barostats are applied [9][10]. The fundamental equation governing MD simulations is: m i d 2 r i d t 2 = − ∇ i U ( r N ) = F i , where F i is the force on atom i , m i is its mass, d 2 r i d t 2 is its acceleration, and U ( r N ) is the total potential energy [10]. This produces time-resolved trajectories that naturally capture dynamical processes such as solvation shell reorganization, diffusion, and conformational transitions. Free Energy Perturbation (FEP) is a well-established method in computational chemistry based on the principles of statistical mechanics. It is designed to compute the free energy difference, ΔF, between two states, A and B, from Molecular Dynamics (MD) or Metropolis Monte Carlo (MC) simulations. The central pillar of FEP is the Zwanzig equation, presented as: Δ F ( A → B ) = F B − F A = − K B T ln 〈 exp ( − E B − E A K B T ) 〉 , where T is the temperature, K B is the Boltzmann constant, and the angular brackets denote an ensemble average over a simulation run for state A [9][18]. This equation provides a link between the microscopic energy fluctuations of a system and its macroscopic thermodynamic properties.

Thermodynamic Integration (TI) is an alternative to FEP that is often more robust for complex transformations [19]. It calculates the free energy difference by integrating the derivative of the Hamiltonian with respect to a coupling parameter, λ, as the system is smoothly transitioned from an initial state (A, where λ = 0) to a final state (B, where λ= 1). With the new potential energy function defined as: U ( λ ) = U ( λ ) + λ ( U B − U A ) . The free energy of this system is defined as: F ( N , V , T , λ ) = − K B T ln 〈 exp ( − U ( λ ) K B T ) 〉 . This approach provides a rigorous thermodynamic path, and the free energy difference is obtained by integrating the ensemble-averaged derivative of the potential energy along this path [9]. Δ F ( A → B ) = ∫ ∂ F ( λ ) ∂ λ d λ = ∫ 〈 ∂ U ( λ ) ∂ λ 〉 d λ = ∫ 〈 U B ( λ ) − U A ( λ ) 〉 d λ [18].

The reliability of a final thermodynamic property prediction derived from first principles is hierarchical, depending critically on the accuracy of the underlying quantum mechanical electronic structure calculation, but also on the fidelity of the subsequent SM treatment used to introduce temperature and entropy effects [12]. This introduces a bottleneck where the error in the final Gibbs free energy may be dominated by the limitations of the SM model, effectively limiting the superior accuracy achieved in the high-level QM electronic energy calculation. This situation necessitates continuous method development to improve the Statistical Mechanical input to match the rigor of the Quantum Mechanical input. In practice, moreover, achieving statistical convergence of ensemble averages remains a critical challenge and a major source of uncertainty, as insufficient sampling of rare but thermodynamically important configurations can dominate the error in computed free energies, independent of—and often comparable in magnitude to—the limitations of the underlying quantum mechanical electronic structure model [20].

Schrodinger’s FEP+ protocol, extensively validated by Wang etal. (2015), demonstrated unprecedented accuracy in predicting ligand binding free energies across diverse protein targets [21]. The study examined 330 small-molecule ligands binding to eight different proteins, including kinases, GPCRs, and proteases, using rigorous free energy perturbation calculations with enhanced sampling techniques. By constructing alchemical transformation pathways that smoothly morphed one ligand into another while maintaining the protein-ligand complex in explicit solvent, they computed relative binding free energies (ΔΔG) with a Root-Mean-Square Error (RMSE) of only 1.1 kcal/mol compared to experimental values, with a correlation coefficient R² = 0.85.

Piana etal. (2012) performed molecular dynamics simulations on Anton to obtain a fully reversible, atomistic description of the folding thermodynamics and kinetics of the 35-residue villin headpiece and multiple mutants [22]. Using the Amber ff99SB*-ILDN force field, each trajectory (≥300 μs) contained between 30 and 150 complete folding/unfolding transitions, constructing the partition function Q ( T ) = ∑ i g i e − β E i , enabling direct computation of Δ G f o l d = − K B T ln ( N f o l d e d / N u n f o l d e d ) , Δ H f o l d , Δ C p , Φ-values, folding rates, and transition-path times. From the folded/unfolded population ratio, they computed folding free energies at multiple temperatures and obtained melting temperatures of 325 K (WT) and 370 K (Nle/Nle), in reasonable agreement with experimental values of 342 K and 361 K. The stability difference between WT and the Nle/Nle double mutant at 360 K (ΔΔG ≈ 2.2 kcal/mol) matched the experimental stabilization of around 1 kcal/mol. Folding enthalpies extracted either from ΔU (force-field energy differences) or van’t Hoff analysis fell in the range 14 - 26 kcal/mol, slightly below calorimetric values (29 kcal/mol), and the computed folding heat capacities were small ( Δ C p 0 - 0.2 kcal/mol∙K⁻¹), substantially lower than experiment (0.46 kcal/mol∙K⁻¹), indicating a systematic force-field deficiency in modeling helix temperature dependence. Sampling also enabled rate extraction, at 360 K, WT villin folded in 16 μs, whereas the Nle/Nle variant folded five times faster (3 μs), matching experimental rate ratios, and simulating mutants in both WT and Nle/Nle backgrounds, they computed Φ-values directly from changes in folding rates and stabilities. The F10L mutation produced a Φ ~ 0.2 in WT—matching experimental Φ-values of 0.3 - 0.6—but Φ = 0.0 in the Nle/Nle background, indicating that the double mutant follows a subtly shifted folding pathway in which helix-3 formation precedes helix-1 packing more strongly than in WT. Finally, an optimized reaction-coordinate analysis yielded free-energy barriers of only 1 - 2 kcal/mol, consistent with microsecond folding and with the transition-path time estimates. The barrier position shifted toward the native basin at higher temperature, exhibiting Hammond behavior. Overall, this work demonstrated that long-timescale MD with TI has matured to the point where folding free energies, enthalpies, heat capacities, rates, Φ-values, and transition-path statistics can be extracted from an atomistic model, achieving broad agreement with experiment—with discrepancies largely confined to Δ C p and absolute folding rates.

Fernández-Serra etal. (2005) compared the performance of two generalized-gradient exchange-correlation functionals—BLYP and RPBE—in abinitio molecular dynamics simulations of room-temperature liquid water, focusing on their ability to reproduce structural and dynamical properties known to be problematic in standard DFT-based AIMD [23]. In the alchemical pathway calculation, the free energy difference was computed as Δ F = ∫ 〈 ∂ U ( λ ) ∂ λ 〉 d λ , where the ensemble average at each λ required the time scale of AIMD simulation at different levels of theory. Using 32-molecule heavy-water trajectories generated with the SIESTA linear-scaling implementation and DZP numerical atomic orbitals, the authors propagated 17 - 30 ps of NVE dynamics following multipicosecond annealing to account for the >20 ps Hydrogen-Bond (HB) network equilibration previously identified as a key source of long-timescale relaxation artifacts. Their analysis of radial distribution functions showed that RPBE yields a systematically less over-structured liquid than BLYP, consistent with a weakened hydrogen bond. Dynamically, RPBE markedly improves the diffusivity, with diffusivity ~ 20% below the experimental value—whereas BLYP underestimates diffusivity by nearly an order of magnitude. Despite the weaker average HB suggested by the elongated O-O distance, RPBE exhibits faster structural relaxation than BLYP—indicating that the functional’s apparent success likely results from error cancellation rather than a fundamentally improved description of water’s underlying energetics. The authors conclude that RPBE provides a practical, computationally inexpensive GGA-level functional whose structural, dynamical, and network properties represent a substantial improvement over BLYP. Resulting from the difference caused by the underlying DFT theory, the accuracy of the final free energy is limited by systematic errors in the energy function E ( r N ) . As abinitio methods approach chemical accuracy, the SM framework propagates error from the underlying energy function, making the QM-SM hierarchy interdependent for first-principles calculations.

The common starting point for most electronic structure calculations is the time-independent, Non-Relativistic Schrödinger Equation (NRSE) [1]. The core component of this equation is the Molecular Hamiltonian (H), an operator that mathematically represents the total energy of the molecular system, with N electrons and M nuclei: H ^ Ψ ( r , R ) = E Ψ ( r , R ) . Where r = { r 1 , r 2 , ⋯ , r n } and R = { R 1 , R 2 , ⋯ , R m } denotes electronic and nuclear coordinates, respectively. The construction of this Hamiltonian reveals the initial conceptual melding of two distinct physical theories, Quantum Mechanics (governing kinetic energy) and Classical Electromagnetism (defining potential energy). The NRSE Hamiltonian can be decomposed into five fundamental terms, representing the major interactions within the system,

with Nuclear Kinetic Energy ( T N ) term, Electronic Kinetic Energy ( T e ) term, Electron-Nucleus Attraction ( V e N ) term, Electron-Electron Repulsion ( V e e ) term, and Nucleus-Nucleus Repulsion ( V N N ) term [18].

The complexity inherent in solving the electronic structure for many interacting electrons, termed the many-body problem, makes achieving an accurate, closed-form analytical solution infeasible for all but the simplest systems, such as the hydrogen molecular ion [1]. The challenge is structurally embedded in the interaction terms themselves. The V e e term, while expressed through a quantum mechanical operator, is derived directly from the classical Coulomb potential. The specific 1/r dependence of the Coulomb potential mandates that every charged particle interacts simultaneously with every other charged particle, regardless of distance. This non-local, long-range nature of the electron-electron interaction forces the development of complex, high-scaling computational approximation methods—like Coupled Cluster theory—required to accurately account for the subtle, instantaneous avoidance behavior known as dynamic electron correlation [4]. The difficulty of quantum chemistry is thus, in a very real sense, a consequence of the foundational constraints imposed by the quantum mechanical Hamiltonian and Classical Electromagnetism.

Specifically, after the choice of a molecular Hamiltonian, the practical solution of the Schrödinger equation requires an additional and unavoidable approximation: the expansion of the many-electron wave function in a finite basis set. In principle, the exact solution is recovered only in the Complete Basis Set (CBS) limit; however, in practice, computational cost necessitates truncation to a finite set of functions, introducing basis-set incompleteness error. This approximation is orthogonal to, yet deeply intertwined with, the hierarchy of electronic structure methods: lower-level approaches such as Hartree-Fock or semi-local density functional approximations often converge more rapidly with basis set size, while systematically improvable, wave-function-based methods such as Møller-Plesset perturbation theory and Coupled Cluster theory require increasingly large and flexible basis sets to fully realize their theoretical accuracy. As a result, the predictive reliability of an abinitio calculation is determined not solely by the formal sophistication of the quantum mechanics-based electronic structure method, but by the balanced convergence of both the many-body treatment and the underlying basis representation of the wave function [11][24][25].

Ruden etal. (2003) examine the quantitative role of connected quadruple excitations by computing CCSDTQ-CCSDT energy differences for a series of small first and second-row molecules, using correlation-consistent basis sets up to cc-pVQZ [24]. The authors determine how these higher-order excitations affect total atomization energies and how their inclusion improves agreement with experimental values. For di-atomics such as N₂, O₂, and F₂, the CCSDTQ correction lies in the range of 3 - 6 kJ/mol, with N₂ receiving about 4 kJ/mol and O₂ approximately 5 - 6 kJ/mol. In contrast, for predominantly dynamical-correlation systems (e.g., hydrides and closed-shell species), quadruple effects are small—typically smaller than 1 kJ/mol. The authors also show that including connected quadruples substantially reduces the residual error relative to experimental atomization energies, especially for multireference di-atomics where lower-level coupled-cluster methods systematically underestimate bond strengths.

In addition, critical assessment of the standard high-level methods was also conducted. CCSD(T)—the prevalent “gold standard”—performs exceptionally well for single-reference molecules, but its perturbative treatment of triples is insufficient for systems with significant non-dynamical correlation, leading to atomization-energy errors of several kJ/mol for N₂, O₂, and F₂. CCSDT also fails to recover the full correlation energy for these challenging molecules, as well as the CCSDTQ does for most cases. Generally speaking, both CCSD(T) and CCSDT leave a consistent, molecule-dependent under-binding that correlates directly with the size of the missing CCSDTQ contribution.

CCSDTQ, although offering clear improvements and bringing theoretical predictions closer to experimental data, suffers from extremely high computational scaling (O(N¹⁰)) and slow basis-set convergence, limiting its applicability to very small systems. In addition, there are cases where CCSD(T) and CCSDT actually perform better than the CCSDTQ method. This steep computational cost prevents the method from becoming routine for large systems. Overall, the paper establishes that CCSDTQ represents a principal source of the remaining deviation between CCSD(T)/CCSDT and experiment in strongly correlated small molecules, while also highlighting the practical computational challenges and lack of generalizability for different molecular systems that prevent CCSDTQ from serving as a general-purpose benchmark method.

The mathematical formalism of relativity consists of two main theories: Special Relativity, which deals with spacetime in the absence of gravity, and General Relativity, which describes gravity as the curvature of spacetime caused by mass and energy. SR is based on two postulates: the laws of physics are the same in all inertial frames and the speed of light in a vacuum, c, is the same for all observers [26]. The fundamental invariant is the spacetime interval, d s 2 = − ( c d t ) 2 + d x 2 + d y 2 + d z 2 , which usually uses a (−, +, +, +) metric signature [27]. If frame S′ moves at velocity v along the x-axis of frame S, the coordinates transform as, c t ′ = γ ( c t − β x ) , x ′ = γ ( x − β c t ) , y ′ = y , z ′ = z , where β = v / c and γ = 1 / ( 1 − β 2 ) [26]. To make laws manifestly invariant, we can use 4-vectors: X μ = ( c t , X , Y , Z ) , V μ = γ ( c , v ) , P μ = ( E / c , p ) , and the Minkowski metric η μ ν = ( − 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 ) , the spacetime interval becomes d s 2 = η μ ν d x μ d x ν [27]. GR’s core idea, inspired by the equivalence principle, is that mass and energy curve spacetime [28]. In GR, spacetime is not flat but a 4-dimensional Lorentzian manifold. The Minkowski metric η μ ν is replaced by a general metric tensor g μ ν ( x ) that can vary from point to point, d s 2 = g μ ν ( x ) d x μ d x ν [29]. The Einstein Field Equation is the heart of GE, G μ ν = 8 π G C 4 T μ ν [30]. For chemical systems, gravitational effects are negligible, and the primary focus is Special Relativity.

Relativity becomes a mandatory input for abinitio theory when dealing with chemical systems containing heavy elements (high atomic number, Z). In these systems, electrons near the massive nucleus experience strong electric fields, causing them to accelerate to velocities comparable to the speed of light (c). Under these conditions, the non-relativistic description derived from the NRSE breaks down. Relativistic Quantum Mechanics (RQM), based on the principles of Special Relativity (SR), provides the necessary framework. RQM formalisms, most famously the Dirac equation, offer a unified, Poincaré-covariant description of quantum mechanics [6]. H ^ D = c α ⋅ p ^ + β ′ m e c 2 + V ( r ) , where p ^ is the momentum operator α and β are Dirac matrices, satisfying the anticommutation relations and c is the speed of light. Crucially, the Dirac equation automatically yields fundamental physical predictions—such as the existence of antimatter, the spin magnetic moments of fermions, and fine structure effects—which, by contrast, must be artificially added to the Hamiltonian operator in non-relativistic quantum mechanics to achieve agreement with experimental observations. Moreover, the explicit treatment of all electrons becomes computationally demanding due to the large number of core electrons and the expensive relativistic effect corrections. The Effective Core Potentials (ECPs) method can be used to replace the inner-shell electrons with an angular-momentum-dependent pseudopotential that reproduces their average influence on the chemically active valence electrons [11].

Scalar Relativistic Effects are spin-independent corrections, such as the mass-velocity term (which increases the electron mass and hence contracts inner orbitals) and the Darwin term (a correction to the potential energy). These effects significantly influence orbital sizes, energies, and consequently, bonding geometries and reaction energetics. Spin-Orbit Coupling (SOC) is the dominant spin-dependent relativistic effect [5]. It describes the magnetic interaction between an electron’s spin and its angular momentum due to motion around the nucleus. SOC causes the splitting of energy levels, is crucial for modeling transition metal complexes and heavy atoms, and governs mechanisms like intersystem crossing. It must be accounted for to accurately model magnetic properties, such as Hyperfine Coupling (HFC), especially the interaction between an unpaired electron spin and a nuclear spin.

The most rigorous abinitio methods that incorporate relativity are known as four-component methods, such as the Dirac-Hartree-Fock (DHF) approach. This is expressed as a relativistic Hamiltonian applied to a many-body system, H ^ D H F = ∑ i = 1 N [ c α i ⋅ p ^ i + ( β ′ i − I 4 ) m e c 2 + V n u c ( r i ) ] + ∑ i < j 1 r i j , where p ^ is the momentum operator α and β are Dirac matrices, c is the speed of light, I 4 is the identity matrix, V n u c ( r i ) is nuclear attraction potential and ∑ i < j 1 r i j is the electron-electron repulsion term. These methods directly solve the Dirac equation, naturally including both scalar and spin-orbit coupling effects [6]. The inclusion of relativistic effects is vital for calculations involving f-block elements, where traditional NRSE approaches fail to capture transitions in the f and d shells accurately. The non-relativistic kinetic energy operator, T ^ e = p 2 / 2 m e , must be replaced by the more complex kinetic term derived from the Dirac Hamiltonian. When Z is large, the deviation between the NR and RQM kinetic energy is significant. Thus, for heavy-element chemistry, starting with a relativistic Hamiltonian derived from RQM is a prerequisite for a legitimate first-principles calculation.

Pyykkö, in his articles, has rigorously demonstrated that relativity is not a minor correction but a structural principle governing the abinitio chemistry of heavy elements and in systems beyond the fifth row—and decisively for Z ≳ 70—the breakdown of the non-relativistic Schrödinger framework makes the four-component Dirac formalism indispensable [5][31]. The relativistic formalism, most completely embodied in the Dirac equation, provides a unified framework that naturally incorporates physical phenomena—such as spin and fine structure—that must be added adhoc to the non-relativistic Hamiltonian. In this picture, the leading scalar-relativistic terms (mass-velocity and Darwin) contract and stabilize s and p1/2 shells, while simultaneously destabilizing and expanding d and f shells, effects that all grow approximately as Z 2 . These orbital distortions, dramatic for the gold atom whose 6s contraction peaks within the periodic table, propagate into molecular structure and explain long-standing anomalies such as the nobility of Au and Hg, and the inert-pair effect in Tl-Bi.

Pyykkö also demonstrated that relativistic corrections to molecular bond lengths—first observed in DF-OCE calculations on AuH and PbH₄—are substantial, typically several picometers. Perturbative and pseudopotential analyses confirmed that the contraction does not simply mirror the shrinking valence orbitals but arises predominantly from large core-core first-order contributions that soften short-range repulsion and shift the potential energy curve. For gold systems, these effects are so pronounced that Au-H bonds shorten by ~7 pm and Au-Au dimer bond lengths contract by nearly 20% with a significant bond energy strengthening compared to a non-relativistic calculation. In conclusion, Pyykkö’s work showed that phenomena such as bond-length contraction, shifts in ionization potentials and electron affinities, anomalous ground-state configurations (5d⁹6s² → 5d¹⁰6s¹ for Pt), and even the magnitude of the contraction cannot be consistently understood without explicit relativistic treatment.

While two-component scalar relativistic approximations can capture these dominant effects at a reduced computational cost, the work of Saue and Visscher has established that the full four-component Dirac-Hartree-Fock (DHF) approach, with its explicit treatment of spin-orbit coupling, is formally rigorous for properties like molecular magnetism and electronic spectra in heavy-element complexes [32]. The practical implementation of these relativistic methods, however, introduces significant computational bottlenecks and places severe constraints on the system sizes and basis sets that can be employed. As detailed by Visscher and Dyall, the four-component DHF method incurs a substantial computational cost, often many times more expensive than its non-relativistic counterpart, due to the complexity of the Dirac matrices and the need for spin-orbit Hamiltonian [33]. Furthermore, the accuracy of the final result is limited not just by the relativistic treatment but by the description of electron correlation. For instance, studies by Liu et al. illustrated that for molecules, achieving quantitative accuracy for excitation energies requires both a four-component relativistic Hamiltonian and a high-level correlated method like Multireference Configuration Interaction (MRCI), a combination that remains prohibitively expensive for all but the smallest molecular systems [6][34]. This hierarchy of approximations—from scalar relativistic density functional theory to fully correlated four-component methods—creates a precision-cost trade-off in this quantum relativistic simulation field as well. These findings collectively confirm that for chemical systems involving elements beyond the 5th period, the non-relativistic Hamiltonian loses its claim, as it neglects the mandatory physical law of special relativity that governs the behavior of core electrons, thereby making relativistic approximation methods a prerequisite for predictive quantum chemistry in heavy-element systems, each with its computational cost, accuracy and specialization.

Electromagnetism provides the physical foundation for the potential energy terms in the molecular Hamiltonian. The Coulomb interaction is responsible for all attractive forces (electron-nucleus) and repulsive forces (electron-electron and nucleus-nucleus). In the non-relativistic limit, these terms define the structure and stability of the molecule [1]. Classical electromagnetism is summarized by Maxwell’s equations: ∇   ⋅   E = ρ ϵ 0 , ∇   ⋅   B = 0 , ∇   ×   E   =   −   ∂ B ∂ t , ∇   ×   B = μ 0 J   +   μ 0 ϵ 0   ∂ E ∂ t , where E is the electric field vector, B is the magnetic field vector, ρ is the electric charge density, J is the electric current density vector, ϵ 0 is the permittivity and μ 0 is the permeability [35]. The two Maxwell equations ( ∇   ⋅   B   =   0 , ∇   ×   E   =   −   ∂ B ∂ t ) are automatically satisfied if we define the fields in terms of potentials. The Scalar Potential ( ϕ ) and Vector Potential (A) are defined as B = ∇ × A , E = − ∇ ϕ − ∂ A ∂ t [36]. The Lorentz invariant and Coulomb Gauge are conserved by the fact that ∇ ⋅ A + 1 c 2 ∂ ϕ ∂ t = 0 and ∇ ⋅ A = 0 [37]. The Relativistic Formulation (Covariant Form) is expressed in x μ = ( c t , x , y , z ) , ∂ μ = ( 1 c 2 ∂ ϕ ∂ t , ∇ ) , J μ = ( c ρ , J x , J y , J z ) , A μ = ( ϕ / c , A x , A y , A z ) , and the six components of E and B are combined into an antisymmetric, 4 × 4 matrix F μ ν = ∂ μ A ν − ∂ ν A μ [38]. The fundamental equations of electromagnetism can be derived from the Principle of Least Action, L E M = − 1 4 μ 0 F μ ν F μ ν [19]. Furthermore, the Lagrangian Density for a charged particle interacting with the EM field can be written as L = L P a r t i c l e + L I n t e r a c t i o n + L E M , and this formalism is the direct bridge to Quantum Electrodynamics (QED), where the classical field A μ is promoted to a quantum operator, and the Lagrangian is used in the path integral to calculate probabilities [39][40].

Beyond defining static interactions, EM principles underpin the calculation of dynamic molecular properties via response theory [4]. This formalism aims to quantify how a quantum system reacts to external influences, such as applied electric and magnetic fields. In quantum chemistry, response theory computes observables like frequency-dependent polarizabilities and hyperpolarizabilities by evaluating energy derivatives with respect to the applied field strength. This capability allows abinitio methods to predict how molecules absorb, scatter, or transmit light, providing a crucial bridge between theoretical electronic structure and experimental spectroscopy. Response theory calculates properties like the dipole polarizability by expanding the electronic energy in a weak, static electric field, E ( F ) = E ( 0 ) − μ α F α − 1 2 α α β F α F β and the polarizability tensor is obtained as the second derivative of the energy with respect to the field, α α β = − ∂ 2 E ( F ) ∂ F α ∂ F β | F = 0 .

Standard electronic structure theory implicitly assumes a classical electromagnetic field (Coulomb interaction). However, to achieve rigor first-principles description in confined or specialized environments, such as molecules placed inside optical cavities, the EM field must be treated quantum mechanically. This transition leads to the formulation of Quantum Electrodynamics (QED) chemistry [13]. For non-relativistic systems coupled to a radiation field, the general Hamiltonian includes terms that account for the kinetic energy of the charges, the energy of the quantized electromagnetic field itself, and various interaction terms, including those linear and quadratic in the polarization fields. The full abinitio QED Hamiltonian (e.g., the Pauli-Fierz Hamiltonian) explicitly includes the quantized electromagnetic field (photons) as dynamical degrees of freedom alongside the electrons and nuclei [8]. H ^ P F = ∑ i 1 2 m e ( p i − q e A ( r i ) ) 2 + V c o l u m n + 1 2 ∫ d r 3 ( ϵ 0 E ⊥ 2 + 1 μ 0 B 2 ) , where A is the vector potential operator for the transverse photon field, and E and B are the transverse electric and magnetic fields. Furthermore, the photon field A can be quantized and expanded as: A ( r ) = ∑ k , λ ℏ 2 ϵ 0 ω k V ( a ^ k λ ϵ k λ e i k r + h . c . ) . The choice of Hamiltonian and gauge is crucial for physical consistency when developing methods to handle quantized fields. Standard quantum chemistry implicitly uses the Coulomb gauge, which corresponds to a specific form of the generalized Hamiltonian under certain conditions. For instance, the Pauli-Fierz Hamiltonian is particularly advantageous because it is bounded from below, allowing energy minimization principles to determine a ground state for the coupled light-matter system. Applying the Power-Zienau-Woolley (PZW) transformation, the Coulomb-gauge Hamiltonian can be expressed in multipolar form: H ^ M P = ∑ j 1 m j [ p j − q j A ( r j ) ] 2 + V c o l u m n + H E M , where p j is the momentum of particle j with charge q j , A ( r ) is the transverse vector potential of the EM field, V c o l u m n is the Coulomb interaction between charges and H E M is the free electromagnetic field Hamiltonian [41]-[43]. These developments ensure that the theoretical framework remains grounded in first principles even when treating highly complex quantum light-matter phenomena.

As mentioned above, rigorous theories have shown that by taking derivatives of the energy with respect to externally applied electric and magnetic fields, abinitio methods can compute frequency-dependent properties such as polarizabilities (α) and hyperpolarizabilities [4]. Helgaker, Coriani, and co-workers (2012) have extensively demonstrated that the principles of the quantization of electromagnetism provide not only the potential energy terms for the molecular Hamiltonian but also the foundational framework for predicting spectroscopic observables through response theory [44]. They systematically reviewed the molecular electronic Breit-Pauli Hamiltonian, incorporating the quantum relativistic corrections, and demonstrated that by formulating molecular response theory within a flexible quasi-energy framework, one can derive rigorous expressions for linear, quadratic, and cubic response functions, from which a vast spectrum of molecular properties can be obtained. For instance, the linear response function yields frequency-dependent properties such as dipole polarizabilities, while its poles directly provide electronic excitation energies. Detailed implementation of this formalism for approximate wave functions, including variational methods like Hartree-Fock and MCSCF theory, and non-variational methods like coupled-cluster theory, was also summarized, where a Lagrangian-based approach is essential. The computational cost for such a calculation is high; however, the accuracy of these predictions cannot be achieved without QFT and Relativity incorporation in those frontier areas.

Work pioneered by Flick, Ruggenthaler, and their collaborators moves beyond this semi-classical picture by treating the electromagnetic field quantum mechanically, a necessity for molecules in optical cavities [45][46]. In Yang etal.’s foundational work on Quantum Electrodynamical Time-Dependent Density Functional Theory (QEDFT), they employ the Pauli-Fierz Hamiltonian to explicitly include the quantized photon field as a dynamical variable [47]. This approach correctly predicts emerging phenomena such as the vacuum Rabi splitting and the modification of chemical reaction rates in cavities. The practical implementation of such QED methods, however, introduces profound new challenges. As discussed by Flick etal., the choice of gauge is crucial for reliable QED-based electronic structure calculations. While the Pauli-Fierz Hamiltonian in the Coulomb gauge provides a rigorously bounded-from-below formulation, the corresponding length-gauge representation is often numerically more stable and better suited for practical quantum-chemical implementations [45][47]. Furthermore, the computational cost escalates significantly as one must now account for the photon degrees of freedom, and the accuracy is sensitive to the description of the electron-photon correlation, an area where approximate functionals are still under active development. The research of these groups collectively underscores that while the classical Coulomb potential suffices for traditional spectroscopy, a true first-principles description of light-matter interactions in quantum environments mandates a transition to a Hamiltonian that unifies non-relativistic matter with a quantized electromagnetic field, thereby completing the hierarchy of abinitio theories.

Quantum Field Theory (QFT) is a theoretical framework in physics that successfully combines Quantum Mechanics and Special Relativity, providing the rigorous formalism necessary for modeling quantum many-body physics. QFT provides the mathematical and conceptual language underpinning the most accurate abinitio wave function methods [4]. The central idea of Quantum Field Theory is to quantize the field itself, with the field ϕ ( x → , t ) becoming a field operator [48]. The term “second quantization” refers to the specific formalism used in Quantum Field Theory (QFT) to describe systems with an arbitrary number of particles by quantizing classical fields [49]. In this framework, a creation operator ( a † , or ψ † ( x ) ) adds a particle to a specific quantum state, and an annihilation operator (a, or ψ ( x ) ) removes a particle from a specific quantum state, with Fock Space as the state space where second quantization operates [50]. For bosons, the operators satisfy commutation relations [ ψ ( x ) , ψ † ( y ) ] = δ ( x − y ) , and for fermions, the operators satisfy anticommutation relations { ψ ( x ) , ψ † ( y ) } = δ ( x − y ) [48][51]. The scalar field can be expanded as ϕ ( x ) = ∫ d 3 k ( 2 π ) 3 2 E k ( a ^ k e i k x + h . c . ) and the Dirac field can be expanded as ψ ( x ) = ∑ s ∫ d 3 p ( 2 π ) 3 2 E p ( a ^ k , s u s ( p ) e i p x + h . c . ) . The Hamiltonian in the second quantized picture can be expressed by the creation and annihilation operators in this case as: H ^ Q F T   =   ∑ k ( a ^ k † a ^ k   + 1 2 ) ℏ ω . The Lagrangian becomes a spatial integral of a Lagrangian density L, L = ∫ d x 3 L ( ϕ , ∂ μ ϕ ) , with the action as S = ∫ d x 4   L [52]. A profoundly different approach is the Path Integral Formulation, where the central object is the path integral or the functional integral, instead of operators and commutation relations [53]. The transition amplitude between two states is given by a sum over all possible field configurations weighted by the classical action, 〈 ϕ f | e − i H ^ t | ϕ i 〉 = ∫ D ϕ   e i S ( ϕ ) / h , where D ϕ is a measure on the infinite-dimensional space of all field configurations [53][54]. In addition, fields integrated over all possible configurations can be obtained by 〈 0 | T { ϕ ( x 1 ) ⋯ ϕ ( x n ) } | 0 〉 = ∫ D ϕ   ϕ ( x 1 ) ⋯ ϕ ( x n ) e i ∫ d x 4   L ∫ D ϕ   e i ∫ d x 4   L . The QFT technique of Second Quantization (SQ) also provides pictures for various molecular electronic structure theories. SQ replaces the standard wave function with an algebraic framework using creation and annihilation operators. The electronic Hamiltonian becomes, H ^   =   ∑ p q h p q a ^ p † a ^ q   + 1 4 ∑ p q r s 〈 p q | | r s 〉 a ^ p † a ^ q † a ^ s a ^ r , where h p q are one-electron integrals and 〈 p q | | r s 〉 are anti-symmetrized two-electron integrals [1][4]. This formalism naturally handles the varying number of particles involved in correlation effects. Coupled Cluster (CC) theory, often cited as the “gold standard” of quantum chemistry (specifically CCSD(T) for single-reference problems), is constructed using SQ [24]. The CC wave function employs an exponential ansatz, | Ψ C C 〉 = e T ^ | ϕ 0 〉 , where T is the cluster operator containing single and double excitation operators T ^ 1 and T ^ 2 is defined using SQ notation. The algebra of SQ and related diagrammatic methods allow for the derivation of algebraic equations for the energy and cluster amplitudes [4][55]. Similarly, Many-Body Perturbation Theory (MBPT), such as Møller-Plesset methods, which describes corrections to a simpler mean-field solution as an asymptotic series, is formalized and generalized within the QFT framework.

Bartlett and Musiał (2007) have demonstrated that the formalism of Quantum Field Theory (QFT), through the technique of second quantization, provides the mathematical foundation for modern abinitio wave function approximation methods at expensive computational costs [55]. In their benchmark study, Bartlett and co-workers showed that CCSD(T)/CBS calculations yield a result accurate enough that they have served as a gold standard for validating more approximate methods [55]. However, this accuracy comes at a formidable O(N⁷) computational cost. Sherrill etal. (2009) underscored the prohibitive computational scaling of canonical CCSD(T), showing that even prototype noncovalent systems such as the benzene dimer approach the practical limits of O(N⁷) cost, and emphasized the necessity of more efficient correlation treatments for quantitatively reliable interaction energies [56]. Another limitation is its failure for strongly correlated systems; as demonstrated by Piecuch and co-workers, the CCSD(T) description of the symmetric bond breaking in the N₂ molecule deviates from the full configuration interaction limit, underscoring the necessity of more complex multireference formulations [57].

The group of Alavi and Booth pioneered Full Configuration Interaction Quantum Monte Carlo (FCIQMC), which stochastically solves the many-body problem. FCIQMC works within the second-quantized form of the Hamiltonian. Shepherd etal. (2012) applied FCIQMC to obtain near-exact ground-state energies for up to 54 electrons, demonstrating that stochastic sampling of over 10¹⁰⁰ Slater determinants in a plane-wave basis can achieve benchmark accuracy and provide a reliable reference for finite-basis and complete-basis extrapolations in periodic systems [58]. Similarly, the Density Matrix Renormalization Group (DMRG) approach, advanced by Chan and White, has proven powerful for one-dimensional systems; Kurashige and Yanai (2011) developed a second-order perturbation theory based on a DMRG-SCF reference, enabling accurate treatment of both static and dynamic correlation in the chromium dimer (Cr₂). They showed that conventional CASSCF/CASPT2 with smaller active spaces significantly overestimates the binding energy, whereas their DMRG-CASPT2 approach reliably reproduces the dissociation curve [59].

The most obvious application of QFT into abinitio chemistry involves Quantum Electrodynamics (QED) chemistry, which studies molecular systems strongly coupled to quantized electromagnetic fields (photons), often achieved within optical cavities [8]. Integration of QED with DFT is an interesting and new paradigm that requires integrating QFT and RQM simultaneously into the molecular description. However, the Density Functional Theory (DFT) framework faces the unknown exchange-correlation (XC) functional challenge. Density Functional Theory (DFT) represents an alternative abinitio framework based on the Hohenberg-Kohn (HK) theorems, which formally prove that all properties of a system, including the total energy, are unique functionals of the ground-state electron density [15]. The Kohn-Sham (KS) formulation of DFT replaces the complicated interacting system with an auxiliary, non-interacting system that yields the same ground-state density [8]. The rigorous nature of the HK theorems is limited by one overwhelming issue: the exact form of the Exchange-Correlation (EXC) functional, which incorporates all non-classical kinetic and potential effects, remains unknown. Therefore, the application of DFT necessitates the use of approximate XC functionals, while many modern functionals are highly effective, often achieving reasonable accuracy with low computational cost (scaling O(N³) to O(N⁴)), the necessity of approximating EXC introduces a theoretical tension regarding DFT’s “first principles” status [60]. This problem centers on whether modern functionals, especially hybrid functionals, which are optimized for high accuracy in energy prediction, compromise the accurate representation of the electron density itself—the core variable mandated by the HK theorems, and since the goal of the theory is to minimize the energy functional with respect to the density, any approximation to EXC means the true exact solution cannot be reached.

Cohen, Mori-Sánchez, and Yang comprehensively examined the challenges facing Density Functional Theory (DFT), focusing on how approximate exchange-correlation functionals fail for certain chemical systems [60]. The authors identify five major challenges: developing functionals that uniformly outperform the ubiquitous B3LYP functional, improving descriptions of reaction barriers and van der Waals interactions, understanding the significance of different optimization approaches, addressing delocalization error, and tackling static correlation error in systems with near-degenerate states. This is exemplified by the simultaneous failure of all current functionals to correctly describe the two simplest molecules in chemistry—infinitely stretched H 2 + and H₂—with typical errors of 50 - 60 kcal/mol despite achieving 3 - 4 kcal/mol accuracy for standard thermochemistry. The delocalization error causes functionals to give low energies for delocalized densities, leading to errors in stretched molecules, reaction barriers, organic isomerization energies, and band gaps in solids, while static correlation error prevents proper description of multireference systems like dissociating bonds. The authors argue that the integer nature of electrons, naturally present in the many-body wavefunction, must be encoded into density functionals through discontinuous behavior, and that current smooth, differentiable functionals cannot fundamentally resolve strong correlation problems regardless of parametrization. Various proposed solutions are discussed, including range-separated functionals that mix short-range DFT with long-range Hartree-Fock exchange, local hybrid functionals, explicit van der Waals functionals with non-local correlation terms, and functionals incorporating unoccupied orbitals like RPA, though none yet satisfactorily address the core issues. However, this might emphasize a philosophical distinction between DFT and abinitio quantum chemistry: while quantum chemistry seeks different wavefunctions Ψ for each system, DFT seeks one universal functional E x c [ ρ ] that works for all densities, meaning the challenge is concentrated in constructing this universal functional rather than solving individual systems. Ultimately, the authors conclude that innovative functional forms incorporating derivative discontinuities are essential for DFT to advance beyond its current status as a successful approximation method and fulfill its promise as an exact theory capable of describing strong correlation, proper band gaps, and the full complexity of chemical bonding across all systems from simple atoms to complex materials.

Development of more generally applicable functionals was also proposed, like the SCAN (Strongly Constrained and Appropriately Normed), a new meta-generalized gradient approximation (meta-GGA) density functional for electronic structure calculations by Sun, Ruzsinszky, and Perdew, which was tested on solid-state and molecular systems and simultaneously improved the description of covalent, ionic, and hydrogen bonds [61]. SCAN satisfies all 17 known exact theoretical constraints while being calibrated against systems where semi-local methods excel, including rare-gas atoms, non-bonded interactions, etc. The functional employs parameters to distinguish between covalent, metallic, and weak bonding environments. This emphasis on theoretical rigor and design enables SCAN to achieve comparative performance across diverse chemical systems without heavily relying on empirical fitting to molecular data. Essentially, functionals that are strongly guided by known exact constraints tend to perform well for both density and energy, validating the continued search for better functional forms [61]. In addition, the traditional Born-Oppenheimer (BO) approximation, a cornerstone of QM-based quantum chemistry, must be generalized, and researchers are developing frameworks based on a Generalized Born-Huang expansion, which treats nuclei, electrons, and photons simultaneously on equal footing, trying to provide a better first-principles integration with QED chemistry phenomena [3][8].

Tom Banks reformulates Density Functional Theory (DFT) using concepts and methods familiar to quantum field theorists [62]. The central idea is to express the DFT problem in terms of the generating functional of connected Green’s functions for the electron density operator in the Homogeneous Electron Gas (HEG). The density functional F[n] is identified as the Legendre transform of this generating functional. Banks emphasize that the full energy functional in an external potential V e x t can be written as: E [ V e x t ] = F [ n ] + ∫ V e x t ( x ) n ( x ) d x 3 . A key proposal is the use of a large-K expansion and introducing a Hubbard-Stratonovich field σ for the Coulomb interaction and scaling the coupling as g = 2 / K . The action becomes: S [ σ ] = 1 2 ∫ σ ( − ∇ 2 ) σ + β L [ σ − i V e x t ] , where L[χ] is the free energy of non-interacting electrons in potential χ. In the large-K limit, the functional integral leads to a self-consistent equation for the effective potential. At leading order, this yields the Hartree approximation: F [ n ] = F 0 [ n ] + 1 2 K ∫ d x 3 d y 3   n ( x ) 1 ∇ 2 ( x , y ) n ( y ) , where F 0 [ n ] denotes the non-interacting kinetic energy functional. Higher-order corrections are systematically computed via loop diagrams involving functional derivatives of L[χ]. Banks also discusses the relationship between the density functional and the quantum effective action Γ[χ] for the Coulomb photon, suggesting that the latter may be easier to compute in momentum space. He further explores the small-K limit, connections to the Kohn-Sham equations, and the application of the large-K expansion to inhomogeneous phases like the Wigner crystal at low densities. This work inspires field theorists to contribute new analytical approaches to DFT, leveraging techniques from quantum field theory to advance this central tool in condensed matter and quantum chemistry.

Svendsen etal. introduce a theoretical framework that combines Density Functional Theory (DFT) with Macroscopic Quantum Electrodynamics (MQED) to quantitatively describe quantum light-matter interactions in 2D materials, particularly in van der Waals heterostructures [39]. The authors address challenges such as ultra-confined plasmonic modes, optical losses, and the need for a first-principles description of electronic states. The core is the MQED description of the vector potential in the Weyl gauge: A i ( r ) = ℏ π ϵ 0 ∫ d ω ω c 2 ∫ d 3 s   G i l ( r , s , ω ) [ Im ϵ ( s , ω ) ] i l f j ( r , s , ω ) + h . c . , and the use of the Green’s function G i l ( r , s , ω ) , which encodes the electromagnetic response of the environment. The light-matter interaction is treated via the Wigner-Weisskopf model, leading to a coupling kernel that incorporates both the electronic transition matrix elements (from DFT wavefunctions) and the photonic density of states. A key mathematical result is the expression for the transition rate: ∑ v q | 〈 f q ⊗ v q | H i n t | i ⊗ 0 〉 | 2 δ ( ω − ω v q ) = ∬ d r d r ′   j * ( r ) Im G ( r , r ′ , ω ) j ( r ′ ) , where j ( r ) is the transition current density. The framework is validated in the weak-to-strong coupling regime, but its limitations are noted for the ultra-strong coupling regime, where electron dressing by the EM field becomes significant. Overall, this work establishes a practical, abinitio-based approach integrating QFT and DFT for predicting light-matter interactions in realistic systems.

Another prominent example of this synthesis is the Polaritonic Dirac-Hartree-Fock (Pol-DHF) approach. Pol-DHF is derived directly from relativistic QED theory using the Coulomb gauge Hamiltonian and the Dirac formalism [63]. It is specifically designed to simulate ground and excited state properties of heavy transition metal complexes strongly interacting with quantum fields, allowing for the accurate description of Spin-Orbit Coupling (SOC) and field-induced effects. Research using Pol-DHF on metal hydrides has confirmed that the magnitude of relativistic effects (SOC) and polaritonic effects (QED coupling) can be quantitatively comparable. As mentioned in the previous paragraph, Ruggenthaler etal. recently developed the full integration of QFT with quantum chemistry through abinitio Quantum Electrodynamics (QED) principles [8][47]. Flick etal. (2017) demonstrated a key finding by simulating the strong coupling of a molecule to a cavity mode, showing that the energy barrier of a chemical reaction could be altered, effectively modifying reaction rates [45]. Konecny etal. demonstrate that a cavity-mediated spin-orbit-like interaction emerges at the four-component Dirac-Kohn-Sham level, and their benchmark calculations for heavy atoms show that the energy shifts induced by strong light-matter coupling can be of comparable magnitude to relativistic spin-orbit effects [18]. These critical findings underscore that only a unified, QFT-based relativistic quantization framework can provide a truly rigorous, predictive first-principles description for molecules in quantum environments with the rigor necessary for the cutting-edge regime, pushing the boundaries of abinitio theory beyond the static molecular Hamiltonian.

The final accuracy and tractability of any abinitio result depend on a cascade of approximations, rooted in the six fundamental physical theories discussed. Each theoretical input provides both necessary structure and inherent limitations, forming an interdependent ecosystem summarized in Table 1.

Table 1.Mapping fundamental physics to quantum chemical components.

The practical deployment of abinitio methods is governed by a fundamental trade-off between the rigor of the physical formalism used (the treatment of electron correlation) and the computational cost, which typically scales polynomially with the number of basic functions (N) (see Table 2).

Table 2.Hierarchy of abinitio approximations and costs.

The endeavor of abinitio quantum chemistry, when examined through the lens of its foundational physical theories, reveals itself to be an intricate balancing act between competing demands: the aspiration for rigorous first-principles prediction and the pragmatic necessity of computational tractability; the desire to incorporate complete physical laws and the recognition that every theory introduces its own domain of validity and inherent limitations. The analysis presented here illuminates a critical insight: there exists no single, universally applicable abinitio method that can simultaneously incorporate all relevant physics at acceptable computational cost for all chemical systems. Instead, the field is characterized by a hierarchy of methods, each representing strategic decisions about which physical effects to treat rigorously and which to approximate or neglect [1][4][11]. The Hartree-Fock method sacrifices electron correlation for computational efficiency [1]. Density Functional Theory trades the formal rigor of wave function approaches for a dependence on approximate exchange-correlation functionals [60]. Coupled Cluster theory achieves systematic accuracy through size-extensive, QFT-based formalism, but at steep polynomial scaling [55]. Relativistic four-component methods capture the physics demanded by heavy elements at the cost of increased formal complexity and computational burden [6][32].

What emerges most clearly from this synthesis is the profoundly interdependent nature of these approximations. An error introduced at one level of theory can propagate through or be masked by approximations at another level. A highly accurate electronic structure calculation performed within the Born-Oppenheimer approximation will fail to capture non-adiabatic charge transfer [2][3]. A sophisticated post-Hartree-Fock treatment of electron correlation becomes moot if the subsequent statistical mechanical treatment of thermal effects relies on overly rigid models of molecular motion. A non-relativistic calculation of a gold complex, regardless of the sophistication of its correlation treatment, cannot legitimately claim first principles status, as it violates the physical laws governing high-velocity core electrons by not including the relativity effects [5][31]. This hierarchical dependence means that method development must proceed on multiple fronts simultaneously: improving electronic structure theory, refining statistical mechanical models, incorporating relativistic and QED effects, and developing techniques to escape the Born-Oppenheimer approximation where necessary. For systems containing heavy elements, the mandatory incorporation of relativity through frameworks like the Dirac equation ensures that calculations remain grounded in the correct physical laws [6][31]. For molecules in optical cavities or other strong-coupling regimes, the explicit quantization of the electromagnetic field via QED becomes essential, requiring simultaneous treatment of electrons, nuclei, and photons on equal footing [8][47]. Methods such as Polaritonic Dirac Hartree-Fock (Pol-DHF) exemplify this unifying direction, demonstrating that relativistic effects (spin-orbit coupling) and field-matter coupling effects can be of comparable magnitude, necessitating a fully integrated QFT-based relativistic framework for rigorous prediction [63]. This progression raises a fundamental question about the ultimate scope of abinitio chemistry: as we systematically replace classical approximations with quantum treatments, neglect of relativistic effects with proper Dirac formalism, and classical electromagnetic fields with quantized radiation fields, are we approaching a complete description of molecular reality, or merely trading one set of approximations for another, more sophisticated but equally incomplete set? The pragmatic answer is likely the latter. Quantum Field Theory itself requires renormalization to handle infinities. Statistical mechanics relies on ergodic assumptions that may not hold for complex landscapes [12]. Even the most advanced QED treatments currently deployed ignore quantum gravity, higher-order QED corrections, and the weak and strong nuclear forces—effects vanishingly small for chemistry but present nonetheless [13].

Whether it is due to the lack of theoretical integration exactness or the core approximation in some steps causing the error to accumulate in subsequent steps, the universal application of abinitio principles for large-scale and precise prediction is still an unmanageable task. For instance, in the solvation free energy prediction process, the 1 kcal/mol error barrier appears difficult to systematically overcome, even with the theoretically rigorous explicit model equipped with quantum mechanics effects, however, the hybrid or semi-empirical implicit model was found to be more accurate in some cases with significantly fewer computational resources [64]. In QM/MM hybrids, high-precision quantum mechanical theories are applied selectively to a chemically critical subsystem—such as a reactive center, chromophore, or catalytic active site—while the surrounding environment is treated using lower-level classical mechanics [64]. Then, the question becomes: since the error in theoretical rigor and computationally expensive abinitio principles is unavoidable, should we develop alternative computational approaches? Indeed, numerous approximation strategies, empirical or semi-empirical, novel hybrid models, and data-driven machine learning models are emerging as alternatives to abinitio models for higher accuracy and less computational cost for specific applications [65][66]. While machine learning and AI-driven models have emerged as promising alternatives for predicting free energies, kinetics, and other chemical properties with reduced computational cost, it is important to note that these models are generally trained on data generated by abinitio calculations or integrated high-level physical models, thereby anchoring their predictions in the very theoretical frameworks that they seek to approximate. Moreover, these innovations are often limited by the core approximation nature of the algorithm and restricted by the transferability for broader applications, requiring the development of various models with complicated parameters.

Looking forward, the field faces both computational and conceptual challenges. Computationally, the steep scaling of high-accuracy methods remains the primary bottleneck, driving development of linear-scaling approaches such as machine learning surrogates, and quantum computing algorithms specifically designed for electronic structure [4][66]. Conceptually, the integration of multiple physical theories—particularly the simultaneous treatment of relativity, strong electron correlation, and quantized fields—remains an active area of method development, requiring new mathematical frameworks and approximations that preserve the essential physics while remaining tractable [8]. Yet this recognition does not diminish the profound utility and continuing evolution of abinitio methods. The past decades have witnessed remarkable progress: Coupled Cluster theory now achieves sub-kilocalorie accuracy for reaction energies; explicitly correlated methods have dramatically accelerated basis set convergence; Born-Oppenheimer molecular dynamics enables abinitio simulation of reactive processes in condensed phases; relativistic treatments have made heavy-element chemistry predictive rather than merely descriptive [4][5][11][15][24][31]. Each advance represents a hard-won extension of the domain over which first-principles calculation provides reliable guidance. The ultimate value of understanding abinitio chemistry as the synthesis of multiple physical theories is both practical and philosophical. Abinitio chemistry does not operate from a single, perfect foundation but from a carefully constructed edifice of interlocking theories, each validated within its own domain, each contributing essential structure to the whole. The field’s continued vitality depends on recognizing both the extraordinary power of this framework—its ability to predict molecular behavior from fundamental constants—and its inherent limitations, rooted in the approximations we must accept to make the infinite complexity of quantum many-body systems yield to practical calculation [1][4]. In this light, abinitio quantum chemistry represents not a march toward a fixed, final theory, but rather a continuous refinement of our ability to incorporate the laws of physics—from Newton to Dirac to Feynman—into predictive models of molecular reality. It is a discipline always balancing the competing imperatives of rigor and feasibility given by the development of computing technologies, always extending the boundaries of what can be reliably calculated from first principles and toward providing genuinely exhaustive descriptions of nature.

This article is a review of previously published literature and does not involve any new studies with human participants or animals; therefore, ethics approval, informed consent, and clinical trial registration are not applicable. All data supporting the conclusions of this review are derived from previously published sources, as cited in the reference list. Any reproduced figures, tables, or text have been appropriately cited, and permissions from copyright holders have been obtained where required.

This work received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.