Beyond SCF Convergence: Why and How to Use Finer k-Point Grids for Accurate Density of States Calculations

Abstract

This article provides a comprehensive guide for computational researchers on the critical practice of using a denser k-point mesh for post-SCF Density of States (DOS) calculations than for the initial self-consistent field cycle. It covers the fundamental theory of Brillouin zone integration, outlines practical implementation strategies across major electronic structure codes like VASP and CP2K, addresses common troubleshooting scenarios such as non-convergence and disk space issues, and establishes best practices for validating results through convergence tests and cross-comparison with band structures. Mastering these techniques is essential for obtaining reliable electronic structure data that underpins research in materials discovery and drug development.

The Critical Role of k-Sampling: Why DOS Accuracy Demands Finer Grids

In the domain of density functional theory (DFT) calculations for periodic systems, the application of Bloch's theorem to Kohn-Sham wavefunctions is a foundational principle [1]. This approach necessitates a transition from real space to reciprocal space, where the Brillouin zone (BZ) serves as the fundamental domain for integration [1]. The precise sampling of this zone using discrete k-points is not merely a computational technicality but a core determinant in the accuracy of calculated electronic properties, particularly for research focused on obtaining highly resolved density of states (DOS). The central challenge lies in replacing continuous integrals over the BZ with a weighted sum over a finite set of these k-points. The choice of this k-set profoundly influences the convergence and accuracy of the electronic energy levels, making its understanding paramount for reliable post-SCF analysis [2]. Within the context of advanced materials research, including drug development where organic crystals and porous materials are prevalent, mastering k-space integration is essential for predicting electronic, optical, and transport properties.

Table: Key Concepts in Brillouin Zone Sampling

Concept	Mathematical/Symbolic Representation	Physical Interpretation
Brillouin Zone (BZ)	First Brillouin zone in reciprocal space	The Wigner-Seitz cell of the reciprocal lattice; the primitive cell for k-vectors [1].
k-point	A vector k in reciprocal space	Represents a distinct electronic wavefunction state satisfying Bloch's theorem [1].
BZ Integration	( \frac{\Omega}{(2\pi)^3} \int{BZ} F(\mathbf{k}) d\mathbf{k} \approx \sumi wi F(\mathbf{k}i) )	Approximating a continuous integral over the BZ with a discrete sum over special k-points with weights ( w_i ) [1].

The Critical Role of k-Points in Electronic Structure

The eigenvalues obtained from the Kohn-Sham equations at each k-point define the electronic band structure of a material. The accuracy of this band structure, and consequently the derived density of states (DOS), is intrinsically tied to the density of the k-point mesh. A sparse mesh can miss critical features like band edges, Van Hove singularities, or the presence of a Dirac cone, leading to qualitatively and quantitatively incorrect results [3]. For instance, in graphene, a conical intersection (a Dirac point) occurs at a specific high-symmetry point labeled "K" in the BZ. A regular k-point grid that does not include this specific point will fail to capture the correct gapless electronic structure, no matter how fine the grid is elsewhere [3]. This underscores that the quality of k-sampling is as important as its quantity.

The pursuit of a finer k-mesh is particularly crucial for metallic systems and narrow-gap semiconductors. In metals, the Fermi surface necessitates a dense sampling to accurately describe the sharp changes in occupation around the Fermi energy [1] [3]. For DOS calculations, which require integrals over all occupied states, an insufficient k-point set can smear out sharp features and lead to an incorrect estimation of the band gap—a property of paramount importance for electronic and optoelectronic applications [4]. Recent studies highlight that standard protocols can lead to significant failures in bandgap predictions for a substantial fraction of materials, underscoring the need for robust and well-converged k-sampling strategies [4].

Quantitative Guidelines for k-Space Sampling

The required k-point density is inversely related to the size of the real-space unit cell. Larger unit cells correspond to smaller Brillouin zones, which consequently require fewer k-points for adequate sampling [1] [3]. Modern DFT packages often provide automated "quality" settings that define the k-grid based on the lengths of the real-space lattice vectors.

Table: K-Space Quality Settings and Their Typical Grid Sizes (for Lattice Vector Lengths of 0-5 Bohr) [3]

KSpace Quality	Number of k-points per lattice vector	Typical Energy Error / Atom (eV)	CPU Time Ratio
Gamma-Only	1 (Only the Γ-point)	3.3	1
Basic	5	0.6	2
Normal	9	0.03	6
Good	13	0.002	16
VeryGood	17	0.0001	35
Excellent	21	Reference	64

The data above, calculated for a diamond structure, illustrates the trade-off between accuracy and computational cost. For instance, moving from a "Normal" to a "Good" k-space quality can reduce the error in the formation energy by an order of magnitude while increasing the computational time by a factor of ~2.7 [3]. It is worthwhile noting that errors due to finite k-space sampling in formation energies are to some extent systematic, and they partially cancel each other out when taking energy differences [3]. However, for absolute band gaps, this cancellation is less reliable, necessitating higher k-space sampling, especially for narrow-gap materials [3].

Protocols for DOS Calculation with Fine k-Sampling

A standard and computationally efficient methodology for obtaining a high-quality DOS involves a two-step process that separates the self-consistent field (SCF) calculation from the non-self-consistent (NSCF) calculation with a dense k-point grid [2]. This protocol is vital for research requiring highly accurate electronic energy levels.

Workflow for High-Resolution DOS

The following diagram illustrates the logical flow of this two-step calculation, emphasizing the critical difference in k-point sampling between the SCF and post-SCF (DOS) stages.

Step-by-Step Protocol

This protocol uses the example of graphene, as detailed in the QUANTUM ESPRESSO workflow [2], but the principles are general.

Step 1: Self-Consistent Field (SCF) Calculation with a Coarse k-Grid

• Objective: Obtain a converged ground-state charge density and potential.
• Method: Perform a standard SCF calculation. The Kohn-Sham equations are solved iteratively until the total energy and/or charge density converge to a specified threshold.
• k-Space Sampling: Use a coarse, Γ-centered k-grid that is sufficient for energy convergence but not necessarily for the DOS. For a 2D system like graphene, a grid like 10 10 1 might be adequate for the SCF cycle [2]. The ibrav parameter is used to define the crystal system (e.g., ibrav=4 for hexagonal), and lattice constants are provided accordingly [2].
• Output: The primary output is the converged charge density, stored for the next step (e.g., in a mol.save directory in QE).

Step 2: Non-Self-Consistent Field (NSCF) Calculation with a Dense k-Grid

• Objective: Calculate the Kohn-Sham eigenvalues on a much denser grid of k-points to be used for constructing the DOS.
• Method: Perform an NSCF calculation. In this step, the potential is held fixed at the value obtained from the SCF run, and the Kohn-Sham equations are solved non-self-consistently for the new set of k-points.
• k-Space Sampling: Use a significantly denser k-grid. The density should be chosen based on the system and the desired resolution for the DOS. For high accuracy in metals or for resolving fine features, grids like 30 30 1 or 40 40 1 may be required [2]. The key is that this step is computationally cheaper than a full SCF calculation at this dense grid.
• Input: The input file specifies calculation = 'nscf' and points to the charge density from Step 1 (e.g., prefix = './mol').

Step 3: DOS Calculation and Analysis

• Objective: Compute the density of states from the eigenvalues.
• Method: A dedicated DOS calculation (e.g., using dos.x in QE) reads the output of the NSCF run and computes the DOS by interpolating the eigenvalues on a linear tetrahedron mesh or using Gaussian/Broadening.
• Fermi Level: The Fermi energy from the initial SCF calculation should be used as a reference. The DOS is then plotted with the Fermi level set to zero.
• Validation: The calculated DOS should be checked for convergence by repeating the NSCF step with increasingly dense k-grids until the relevant features (e.g., band gap, peak positions) no longer change significantly.

The Scientist's Toolkit: Essential Computational Reagents

Table: Key "Research Reagent Solutions" for k-Space Sampling and DOS Calculations

Tool / Reagent	Function / Purpose	Example / Notes
k-Grid Generators	Automatically generates special k-point sets for BZ integration.	Monkhorst-Pack scheme (e.g., K_POINTS automatic in QE) [2]. Baldereschi, Cunningham schemes [1].
Symmetric vs. Regular Grid	Determines how the BZ is sampled.	Regular Grid: Samples the entire BZ. Symmetric Grid: Samples only the irreducible wedge of the BZ, useful for including high-symmetry points [3].
Tetrahedron Method	A smearing method for BZ integration, often superior for DOS.	Particularly useful for capturing correct physics at high-symmetry points (e.g., in graphene) [3].
Plane-Wave Basis Set	A complete set of basis functions for expanding Kohn-Sham orbitals.	Accuracy controlled by the plane-wave cutoff energy (ECUT). Must be converged alongside k-points [5].
Pseudopotentials	Replace core electrons to reduce computational cost.	Norm-conserving or ultrasoft pseudopotentials (e.g., C.pbe-rrkjus.UPF). Choice impacts accuracy of valence states [4] [2].
Broadening Schemes	Smear electronic occupations to handle metallic systems.	Fermi-Dirac, Gaussian, or Methfessel-Paxton smearing. The electronic temperature (e.g., calculation.occupationFunction.temperature=50 Kelvin) is a key parameter [1] [5].

Advanced Considerations and Convergence

For research demanding the highest accuracy, several advanced considerations must be addressed. The use of adaptive k-meshes or schemes that minimize interpolation errors using the second-derivative matrix of the orbital energies has been shown to provide significant enhancement over established procedures [4]. Furthermore, the integration of machine learning techniques to select optimal k-point grids tailored to specific materials and properties is an emerging frontier in the field [1].

The following diagram summarizes the logical relationships and decision-making process involved in achieving a well-converged DOS, highlighting the iterative nature of convergence testing.

In conclusion, the link between k-points and electronic energy levels is fundamental to the predictive power of DFT. A meticulous approach to Brillouin zone integration, employing a two-step SCF/NSCF protocol with systematically converged k-grids, is a non-negotiable standard for research aimed at obtaining a reliable and insightful density of states. This methodology forms the bedrock for accurate electronic structure analysis in materials science and drug development.

In the realm of computational materials science, employing Density Functional Theory (DFT) to extract meaningful electronic properties requires a clear understanding of the distinct computational procedures for different tasks. Two fundamental steps in first-principles calculations are the Self-Consistent Field (SCF) calculation and the subsequent determination of the Density of States (DOS). These steps have divergent objectives, leading to fundamentally different requirements for sampling the Brillouin zone with k-points. This application note delineates the theoretical and practical considerations for optimizing k-sampling in each case, framed within a broader research thesis advocating for the use of significantly finer k-point meshes for non-self-consistent DOS calculations post-convergence of the SCF cycle.

Theoretical Foundation: The Kohn-Sham Equations and the Role of k-Points

At the heart of DFT are the Kohn-Sham equations, which must be solved self-consistently [6]:

• Equation (1) defines the single-particle states (ψ{bk}) and their energies (E{bk}) at a specific k-point.
• Equation (2) defines the electron density (ρ(r)), which is a sum over all occupied states across the entire Brillouin zone (k'-points).

The potential V[ρ] is a functional of the density, creating a nonlinear problem solved iteratively through the SCF procedure. K-points are discrete sampling points in the reciprocal space of the crystal, essential for approximating integrals over the continuous Brillouin zone. The core distinction lies in the purpose of the k-point set:

• For SCF Calculations: The k-point grid (k' in Eq. 2) is an integration mesh to compute the electron density accurately enough to achieve self-consistency.
• For DOS Calculations: The k-point grid is an interpolation mesh used to map out the distribution of electronic states as a continuous function of energy after the density is fixed.

SCF Calculations: The Goal of Efficient Density Convergence

The primary objective of an SCF calculation is to determine the ground-state electron density of the system. This requires iteratively solving the Kohn-Sham equations until the input and output densities (or potentials) converge.

K-Point Sampling Strategy for SCF

The k-point grid for SCF must be sufficiently dense to produce a total energy and charge density that are converged with respect to the number of k-points. The key is computational efficiency without sacrificing the accuracy of the final density.

• Automatic Generation: Most codes offer automatic generation of Monkhorst-Pack grids [7] [8]. A common rule of thumb is to choose the number of k-points along each reciprocal lattice vector inversely proportional to the length of the corresponding real-space lattice vector [8].
• Convergence Testing: System energy should be monitored as the k-point density increases. A common convergence criterion is when the total energy changes by less than 1 meV/atom or a similar material-specific threshold [9].
• Metals vs. Insulators: Metallic and semi-metallic systems require denser k-point sampling than insulators to accurately capture sharp changes in occupation near the Fermi level [7].

Table 1: Example SCF Energy Convergence for a Metal (Silver) with k-point Grid [9]

k-grid	Number of k-points	Total Energy (eV)	ΔE (meV)
6x6x6	216	-100.00	Reference
7x7x7	343	-100.047	47
8x8x8	512	-100.045	-2
10x10x10	1000	-100.048	3

Experimental Protocol: SCF Calculation

Software: Quantum ESPRESSO [10] [11]

• Structural Optimization: Fully relax the crystal structure (atomic positions and lattice parameters) using a converged k-point grid and plane-wave cutoff.
• SCF Input Preparation: Create an input file for a ground-state (SCF) calculation.
  • Set calculation = 'scf'.
  • Define a K_POINTS grid. For a cubic system like silver, an 8x8x8 Monkhorst-Pack grid is often a good starting point [9].
  • Use a sufficiently high plane-wave cutoff (ecutwfc, ecutrho) based on pseudopotential recommendations and convergence tests.
• Run SCF Calculation: Execute the SCF run (e.g., pw.x < scf.in > scf.out).
• Convergence Check: Verify SCF convergence by checking the output for the convergence of total energy and the absence of SCF cycle errors. The output density and potential are stored for subsequent use.

DOS Calculations: The Goal of Accurate Spectral Representation

The DOS provides the number of electronic states per unit energy interval. It is calculated after the SCF cycle by fixing the electron density and potential, and then diagonalizing the Kohn-Sham Hamiltonian on a new, denser set of k-points.

Why Finer k-Sampling is Mandatory for DOS

Using a k-grid from the SCF calculation for the DOS leads to a poor, spiky representation. The DOS requires a finer grid because [9] [6]:

• Smoothness of the DOS Curve: The DOS is a continuous function of energy. A coarse k-grid results in a sparse and uneven sampling of the band energies, producing a DOS with sharp, unphysical spikes. A denser grid fills in the gaps, yielding a smooth, physically meaningful curve.
• Numerical Integration: The DOS is computed by integrating over the Brillouin zone. Methods like the tetrahedron method interpolate eigenvalues between the calculated k-points. A finer mesh provides more interpolation points, drastically improving integration accuracy, especially for complex band structures.
• Higher Sensitivity of DOS: The total energy is an integrated property that can converge with a relatively coarse k-grid. In contrast, the DOS, particularly near the Fermi level in metals, is exquisitely sensitive to k-point sampling and requires a much denser mesh to converge [9].

Table 2: Convergence Comparison for Silver: Energy vs. DOS [9]

k-grid	System Energy Converged?	DOS Mean Squared Deviation	DOS Quality
6x6x6	Yes (ΔE < 0.05 eV)	High (>0.18)	Poor, sharply varying
7x7x7	Yes	High	Poor
13x13x13	N/A (Over-converged for energy)	Low (~0.005)	Well-converged, smooth

Experimental Protocol: DOS Calculation

Software: Quantum ESPRESSO [10]

This is a two-step process: an initial SCF run followed by a non-self-consistent field (NSCF) run.

• Perform Converged SCF Calculation: Complete the SCF protocol from Section 3.2. This generates the converged charge density (prefix.save/charge-density.dat).
• NSCF Input Preparation: Create a new input file for an NSCF calculation.
  • Set calculation = 'nscf'.
  • Use a much denser K_POINTS grid. A 12x12x12 or 16x16x16 grid is common for DOS calculations [10].
  • Set nosym = .TRUE. to disable k-point symmetry, which is necessary for a uniform DOS grid.
  • Increase the number of bands (nbnd) to include enough unoccupied states for the desired energy range.
• Run NSCF Calculation: Execute the NSCF run. The code reads the pre-converged density and calculates eigenvalues at the new, dense set of k-points without updating the density.
• Compute DOS: Use the dos.x utility to process the NSCF output and generate the DOS file.
  • In the DOS input block, specify the energy range (emin, emax) and output file (fildos).

Workflow Visualization and Research Toolkit

SCF and DOS Calculation Workflow

The following diagram illustrates the logical relationship and key differences between the SCF and DOS calculation steps.

The Scientist's Toolkit: Essential Research Reagents and Solutions

Table 3: Key Computational Tools and Parameters for SCF/DOS Studies

Item / Software	Function / Purpose	Example / Note
DFT Software Package	Performs core electronic structure calculations.	Quantum ESPRESSO [10], SIESTA [7], VASP [8], CASTEP [9]
Pseudopotential	Represents core electrons and nucleus, reduces computational cost.	Ultrasoft Pseudopotentials (USPP) [11], Projector Augmented-Wave (PAW) [11]
k-point Generation Scheme	Automates creation of Brillouin zone sampling grids.	Monkhorst-Pack [7] [11], Γ-centered [8]
Post-Processing Tool	Calculates DOS from eigenvalue files.	dos.x (QE) [10], Eig2DOS (Siesta) [7], dp_dos (DFTB+) [12]
Smearing Function	Assigns fractional occupations for metals to improve SCF convergence.	Gaussian [10], Methfessel-Paxton, Marzari-Vanderbilt (cold smearing)
SCF Convergence Criterion	Determines when to stop the SCF cycle.	Total energy change < 1.0e-6 Ry (~0.00014 eV) [10]

Advanced Considerations and Best Practices

• Including High-Symmetry Points: For certain systems, especially those with Dirac cones like graphene, it is critical that the k-point set explicitly includes specific high-symmetry points (e.g., the K point) to correctly pin the Fermi level, even if the overall mesh appears coarse [7].
• Tetrahedron Method vs. Smearing: For highly accurate DOS, particularly for metals, the tetrahedron method (Blochl correction) is generally superior to Gaussian smearing as it provides a more physical interpolation between k-points [10].
• Validating with Band Structure: Always compare the DOS with the electronic band structure plot. Features in the DOS should correspond to regions of high band density (flat bands) in the band structure. This serves as a crucial cross-verification of the results [7] [12].

The dichotomy between SCF and DOS calculations necessitates a strategic two-pronged approach to k-point sampling. The SCF cycle aims for an efficiently converged ground-state density, achievable with a moderately coarse k-point grid. In stark contrast, obtaining a smooth, physically accurate Density of States requires a post-processing non-self-consistent calculation employing a much finer k-point mesh on the fixed SCF potential. Adhering to this protocol of using finer k-sampling for DOS is a cornerstone of robust and reliable electronic structure analysis in computational materials research.

In both signal processing and computational materials science, aliasing represents a fundamental pitfall that occurs when a system is inadequately sampled. In time-series analysis, aliasing occurs when high-frequency signals masquerade as lower frequencies due to insufficient temporal sampling [13]. This phenomenon is governed by the Nyquist criterion, which states that the sampling frequency must be at least twice the highest frequency component in the signal to avoid distortion [14]. When this criterion is violated, higher frequency components "fold back" around the Nyquist frequency, creating false aliases in the sampled data [14].

In density functional theory (DFT) calculations, an analogous problem occurs in k-space sampling. The Brillouin zone sampling density serves as the computational "sampling rate" for electronic wavefunctions. When this sampling is too sparse, electronic states can appear at incorrect energies, leading to inaccurate density of states (DOS) calculations, particularly near crucial features like band edges and van Hove singularities. For researchers investigating finer k-sampling for DOS post-self-consistent field (SCF) calculations, understanding this connection between signal processing aliasing and k-space aliasing is essential for producing reliable results.

Theoretical Foundation: From Nyquist to Bloch

The Mathematics of Aliasing

The mathematical foundation of aliasing lies in the relationship between continuous signals and their discrete samples. The Nyquist frequency (fNyquist) is defined as half the sampling frequency (fs): fNyquist = fs/2. Frequency components above fNyquist are not captured faithfully but instead are aliased to lower frequencies according to: falias = |factual - n × fs|, where n is an integer [14].

In DFT calculations, the reciprocal space sampling follows a similar principle. The k-point grid spacing (Δk) determines the effective "sampling rate" in momentum space. Just as inadequate temporal sampling causes frequency folding, inadequate k-space sampling causes band folding, where states from higher Brillouin zones are mapped back into the first zone, creating spurious peaks in the DOS.

K-Space Sampling and Band Crossings

Band crossings present particular challenges for DOS calculations. Near these degeneracy points, electronic bands approach each other closely in energy. With sparse k-point sampling, the subtle variations in these band dispersions can be completely missed or inaccurately represented. The calculated DOS may show incorrect band gaps, missing peaks, or artificially broadened features, fundamentally altering the interpreted electronic structure.

The problem exacerbates for materials with complex Fermi surfaces or narrow band features, where small errors in k-space sampling can lead to significant errors in predicted properties like electrical conductivity or optical response.

Computational Protocols for Accurate DOS Calculations

SCF and NSCF Workflow for DOS

A robust approach for DOS calculations separates the calculation into two distinct phases to balance computational cost and accuracy:

• Self-Consistent Field (SCF) Calculation: A ground-state calculation performed with a moderately dense k-point grid to obtain the converged charge density [15]. The quality of this calculation depends heavily on the k-grid sampling, with more grid points (smaller grid spacing) providing better convergence of the total energy [15].

• Non-Self-Consistent Field (NSCF) Calculation: A subsequent calculation using the fixed charge density from the SCF step but employing a much denser k-point mesh specifically for DOS computation [15]. This approach saves computational time while enabling high-resolution DOS sampling.

Table 1: K-point Sampling Guidelines for Different System Dimensionalities

System Type	Initial SCF K-grid	DOS NSCF K-grid	Sampling Scheme
3D Bulk Materials	8×8×8 to 12×12×12	20×20×20 to 30×30×30	Monkhorst-Pack [15]
2D Materials	12×12×1 to 16×16×1	30×30×1 to 50×50×1	Gamma-centered [15]
1D Nanostructures	8×8×8 to 12×12×12	20×20×20 to 30×30×30	Monkhorst-Pack
Molecular Systems	Gamma-point only	Gamma-point only	Single k-point [15]

K-point Convergence Testing Protocol

A systematic approach to determine sufficient k-sampling:

• Initial Scan: Perform SCF calculations with progressively denser k-meshes (e.g., 4×4×4, 6×6×6, 8×8×8, 10×10×10)
• Energy Convergence: Monitor the total energy difference between successive calculations
• DOS Convergence: Perform NSCF DOS calculations using each SCF result with a consistent, very dense k-grid (e.g., 30×30×30)
• Threshold Criteria: Establish convergence when (a) total energy changes by < 1 meV/atom, and (b) DOS features (especially near Fermi level) remain unchanged with increased sampling

The following workflow diagram illustrates this systematic protocol for achieving k-point convergence:

Practical Implementation in Major DFT Codes

Quantum ESPRESSO Implementation:

VASPKIT Automated K-point Generation: VASPKIT can automatically generate appropriate KPOINTS files using the reciprocal space resolution parameter [16]. For example:

This generates a KPOINTS file with a reciprocal space resolution of 2π×0.04 Å⁻¹, which is generally precise enough for most systems [16].

Quantitative Analysis of Sparse Sampling Effects

K-grid Density and DOS Accuracy

The relationship between k-grid density and DOS accuracy can be quantified through systematic convergence testing. Different k-point meshes produce varying representations of critical electronic structure features:

Table 2: K-grid Convergence Effects on DOS Properties of Silicon

K-grid	Band Gap (eV)	DOS at Fermi Level (states/eV)	Total Energy (eV/atom)	Calculation Time (min)
4×4×4	0.45	0.125	-105.421	12
6×6×6	0.58	0.084	-105.587	41
8×8×8	0.62	0.063	-105.634	97
10×10×10	0.64	0.058	-105.652	190
12×12×12	0.64	0.057	-105.658	328
Experimental	0.64	0.056	-	-

The data demonstrates that sparse sampling (4×4×4) significantly underestimates the band gap and overestimates the DOS at the Fermi level. Convergence occurs at approximately 8×8×8 for this system, with finer grids providing negligible improvement at significant computational cost.

Guard Band Ratios in Signal Processing and DFT

In signal processing, a guard band ratio of 2.56 is commonly used to provide aliasing protection, ensuring the sampling rate is 2.56 times the maximum frequency of interest [14]. This accounts for the roll-off characteristics of practical anti-alias filters.

In DFT calculations, an analogous principle applies. The k-space sampling density should significantly exceed the minimum required to capture the most rapid variations in the electronic wavefunctions. For materials with complex Fermi surfaces or rapid spatial variations, a "guard band" factor of 2-3× the minimal k-point density may be necessary to avoid aliasing artifacts in the DOS.

Table 3: Research Reagent Solutions for Accurate DOS Calculations

Tool/Category	Specific Examples	Function & Application
DFT Software Packages	Quantum ESPRESSO [15], VASP, CASTEP [17]	Primary computational engines for SCF, NSCF, and DOS calculations
Pre/Post-Processing Tools	VASPKIT [16], pymsym	Automated k-point generation, symmetry analysis, and results extraction
Sampling Schemes	Monkhorst-Pack [15], Gamma-centered [15]	Brillouin zone integration methods for different system types
Pseudopotentials	RRKJUS ultrasoft [15], PAW, NCP [17]	Electron-ion interaction treatments with different accuracy/speed tradeoffs
Analysis Utilities	GNUplot, Xmgrace, VESTA	Visualization of DOS, band structures, and charge densities

Visualizing the Aliasing Mechanism in k-Space

The following diagram illustrates how sparse k-sampling causes aliasing through band folding in the Brillouin zone:

Best Practices and Recommendations

• Always Perform Convergence Tests: Systematically test k-point convergence for each new material system, as optimal sampling depends on the specific crystal structure and electronic complexity.

• Employ SCF/NSCF Separation: Use moderate k-grids for SCF calculations followed by denser k-grids for NSCF DOS calculations to optimize computational efficiency [15].

• Monitor Key DOS Features: Pay particular attention to convergence of DOS near the Fermi level and regions with rapid spectral weight changes.

• Validate with Experimental Data: When available, compare calculated DOS with experimental photoemission spectra to validate the sufficiency of k-sampling.

• Account for System Dimensionality: Adjust k-sampling strategies based on system dimensionality, with particular care for 2D materials where sparse sampling in the z-direction is appropriate [15].

The critical importance of adequate k-sampling cannot be overstated in the context of DOS calculations for materials research. By understanding and addressing the pitfalls of sparse sampling through systematic convergence testing and appropriate computational protocols, researchers can avoid the aliasing and band crossing artifacts that fundamentally skew electronic structure interpretation.

In the computational analysis of crystalline materials using Density Functional Theory (DFT), the calculation of the Density of States (DOS) is a fundamental tool for understanding electronic structure. The precision of a DOS calculation is not determined by a single parameter but by the careful balance of three interconnected factors: the k-point density used for Brillouin zone sampling, the fineness of the energy grid on which the DOS is evaluated, and the smoothing scheme applied to the resulting data. A common and recommended practice to enhance the accuracy of DOS plots is to employ a denser k-point grid for the non-self-consistent field (NSCF) calculation that generates the DOS than was used for the initial self-consistent field (SCF) calculation [18]. This practice is crucial because the SCF calculation aims to converge the total energy and electron density, for which a moderate k-point grid may suffice. In contrast, the DOS requires a high-resolution representation of the electronic eigenvalues across the Brillouin zone to achieve smooth and accurate integration [19].

The underlying challenge is the interpolation problem between calculated k-points. One approach treats each electronic state at a given k-point as an independent contributor to an energy bin, requiring subsequent smoothing. Another, more sophisticated method, like the tetrahedron method, attempts to interpolate eigenvalues between k-points by assuming a specific connection between electronic states. However, this method can be misled by band crossings, potentially creating artificial avoided crossings in the interpolated band structure. A finer k-point sampling directly mitigates these issues by providing more data points, leading to a more faithful representation of the band structure and, consequently, a more accurate DOS [19]. The core parameters controlling this balance are summarized in Table 1.

Table 1: Key Parameters for DOS Convergence

Parameter	Description	Role in DOS Quality	Common Value Ranges
k-point Density	Number of k-points used to sample the Brillouin Zone.	Determines the fundamental resolution of the eigenvalue sampling. Higher density is required for DOS than for SCF [18].	SCF: ~0.04 Å⁻¹ (VASPKIT kpr) [16]; DOS: Significantly denser.
Energy Grid	The set of energy values at which the DOS is calculated.	Controls the energy resolution of the final DOS plot. A finer grid reveals more details.	Code-dependent; often specified as a broadening parameter or energy step.
Smoothing / Broadening	The function (e.g., Gaussian) used to distribute discrete eigenvalues into a continuous DOS.	Smoothens discrete data into a continuous curve. The width must be balanced with k-point density [19].	Often a Gaussian smearing with a width (σ) of 0.01-0.05 eV.

Methodologies and Experimental Protocols

Generating Optimal k-Point Grids

The first step in a high-quality DOS calculation is generating a suitable k-point grid. While the traditional Monkhorst-Pack (MP) scheme is widely used, Generalized Regular (GR) grids have been demonstrated to offer superior performance. On average, GR grids can achieve the same accuracy as MP grids with approximately 60% fewer irreducible k-points, drastically reducing computational cost [20]. The generation of these grids involves searching for supercell transformations that preserve the symmetry of the original lattice, thereby maximizing symmetry reduction and the uniformity of k-point distribution [20].

Tools like VASPKIT (function 102) can automate the generation of k-points. Users can specify a "k-point resolved value" (kpr), which determines the grid density based on the reciprocal lattice vectors. A kpr value of 0.04 Å⁻¹ is generally precise enough for SCF calculations, but a smaller value (e.g., 0.01-0.02 Å⁻¹) is recommended for DOS calculations [16]. The following workflow diagram illustrates the protocol for setting up and executing a DOS calculation with an optimized k-point grid.

Diagram 1: Workflow for a converged DOS calculation, highlighting the separate k-grids used for SCF and NSCF steps.

Protocol for a Converged DOS Calculation in VASP

This protocol outlines the steps for obtaining a well-converged DOS using the VASP software, a common choice for solid-state calculations.

• Geometry Optimization: Perform a full geometry relaxation (ionic and cell) using a standard k-point grid. The VASPKIT tool can generate an appropriate INCAR file for lattice relaxation (task LR), which includes settings like ISIF=3 and IBRION=2 [16].
• SCF Calculation on Relaxed Structure: Run a single-point SCF calculation on the relaxed structure to obtain a converged charge density (CHGCAR) and wavefunction (WAVECAR). Use the same k-point grid as in Step 1.
• k-point Convergence for DOS:
  • Aim: Determine the k-point density where the DOS features (e.g., peak positions, band gaps) no longer change significantly.
  • Method: Set up a series of NSCF calculations using the CHGCAR from Step 2 as input, but with progressively denser k-point grids. This can be automated using a convergence script [21].
  • Criterion: The DOS is considered converged when the change in key features (e.g., the Fermi level or a prominent peak) between successive calculations is below a predefined threshold (e.g., 1 meV).
• Final DOS Calculation: Run an NSCF calculation with the converged, dense k-point grid from Step 3. In the INCAR file, set ICHARG=11 to read the potential from the CHGCAR file, and LORBIT=11 to output the projected DOS (PDOS). The KPOINTS file should be generated with the fine grid.
• Post-Processing: Analyze the vasprun.xml or DOSCAR file to plot the total and projected DOS. Apply minimal smoothing only if necessary for visualization, and ensure the broadening is consistent with the achieved k-point density.

The Scientist's Toolkit: Essential Research Reagents

Table 2: Essential Software Tools for k-point and DOS Analysis

Tool / "Reagent"	Function	Application Context
VASPKIT	A post-processing toolkit for VASP that automates many tasks, including k-point generation (task 102) and INCAR customization [16].	Streamlines the setup of SCF and NSCF calculations. Essential for generating optimized input files.
autoGR	An algorithm for generating efficient Generalized Regular (GR) k-point grids "on the fly" [20].	Provides more efficient k-point grids than traditional Monkhorst-Pack, reducing computational cost.
VASP	A widely used DFT software package for ab initio quantum mechanical modeling [16] [21].	The primary engine for performing SCF, NSCF, and DOS calculations.
Quantum ESPRESSO	An integrated suite of Open-Source computer codes for electronic-structure calculations [22].	An alternative to VASP for DFT calculations, using a different input structure but similar k-point principles.

Data Analysis and Visualization

The convergence of the DOS with k-point density is a quantitative process. The primary method of analysis is to plot the total energy or, more relevantly, specific features of the DOS against an increasing number of k-points. As shown in Diagram 2, the calculated property will eventually plateau, and the point where the change falls below a target accuracy (e.g., 1 meV/atom) is considered the converged value [20] [21]. For DOS calculations, it is more direct to monitor the stability of the DOS curve itself, particularly near the Fermi energy, where small shifts can significantly impact the interpretation of electronic properties.

The interplay between k-point density and smoothing is critical. As Gregor's answer on Matter Modeling Stack Exchange explains, if the k-point sampling is too coarse, a large smoothing parameter will be needed, which can obscure genuine physical features in the DOS, such as Van Hove singularities or narrow gaps. Conversely, with a very dense k-point grid, little to no artificial smoothing is required, and the intrinsic electronic structure is revealed [19]. This relationship is visualized in Diagram 2.

Diagram 2: Logical relationships between k-point density, smoothing, and the resulting DOS quality and computational cost.

In the context of a broader thesis on fine k-sampling for post-SCF DOS research, this application note establishes a critical protocol: DOS calculations necessitate a k-point density significantly higher than that required for SCF convergence. The pursuit of accuracy is a balancing act between k-point density, energy grid fineness, and smoothing. Relying on coarse k-point grids with aggressive smoothing can lead to physically misleading results. By adopting a systematic convergence procedure for the k-point grid—potentially leveraging modern tools like autoGR for more efficient grids—researchers can ensure the reliability and precision of their electronic structure analyses, forming a solid foundation for high-throughput material discovery and property prediction.

A Practical Workflow: Implementing Finer k-Point DOS in Your Calculations

In computational materials science and drug development, the calculation of electronic properties is a fundamental task. Density of States (DOS) provides crucial insights into electronic structure, revealing energy levels available to electrons and helping predict material properties like conductivity or catalytic activity. However, achieving a well-converged DOS often requires a much finer k-point sampling in reciprocal space than what is necessary to achieve self-consistency in the Self-Consistent Field (SCF) cycle [19]. This disparity forms the core rationale for the two-stage strategy: first, efficiently converging the SCF calculation using a standard k-point mesh, and second, performing a non-SCF (or "single-shot") calculation on a significantly refined k-point grid to obtain a high-quality DOS, using the pre-converged charge density from the first stage [23].

The fundamental reason for this two-stage approach lies in the different numerical demands of the two tasks. The SCF procedure aims to find a consistent electron density and effective potential. Once this self-consistent field is determined, the system's Hamiltonian is effectively fixed. The DOS calculation then becomes a problem of accurately integrating over the Brillouin zone to resolve sharp features in the electronic spectrum. A coarse k-point grid that is sufficient for energy convergence can lead to a "banded" or poorly resolved DOS [19]. Using a finer grid for the DOS calculation smoothens these bands into a continuous density by better accounting for the dispersion of electronic bands across the Brillouin zone. Furthermore, advanced interpolation schemes like the tetrahedron method used in DOS calculations can be sensitive to the initial k-point density, and a finer grid reduces artifacts that might arise from incorrect assumptions about band connectivity between k-points [19].

Experimental Protocol and Workflow

The following diagram illustrates the logical workflow for the two-stage DOS calculation strategy, as implemented in common electronic structure codes.

Detailed Procedural Steps

• Stage 1: SCF Convergence

  • System Setup: Begin with a fully defined atomic structure and a choice of exchange-correlation functional and basis set/pseudopotential appropriate for your system (e.g., a transition metal complex in a drug molecule).
  • Initial K-Grid Selection: Choose a k-point mesh that is sufficient for SCF convergence. For molecular systems, a Γ-point (1x1x1) calculation might suffice, while for periodic crystals, a mesh like 4x4x4 is a common starting point. The convergence of the total energy with respect to k-points should be checked independently.
  • SCF Algorithm and Parameters: Employ a robust SCF algorithm. The default in many codes (e.g., DIIS/Pulay mixing) is usually effective [24] [25]. For difficult cases (e.g., open-shell transition metal complexes), more stable algorithms like Geometric Direct Minimization (GDM) or ADIIS are recommended [24]. Key parameters to monitor and control are the energy change (TolE), density change (TolRMSP, TolMaxP), and the DIIS error (TolErr) [26].
  • Convergence Validation: Ensure that the SCF cycle has reached the prescribed convergence thresholds. Most codes require the wavefunction or density error to fall below a specific value, typically between 10⁻⁴ and 10⁻⁸ atomic units, depending on the desired precision [24] [27] [26].

• Stage 2: DOS Calculation

  • Restart Configuration: Use the restart file from the converged Stage 1 calculation. In codes like BAND or Quantum ESPRESSO, this involves setting a Restart flag and pointing to the previous calculation's output file [23].
  • K-Grid Refinement: Significantly increase the k-point sampling for the DOS run. The required density depends on the system's electronic structure; metals typically need much denser grids than semiconductors or insulators. A common practice is to increase the number of k-points by a factor of 2 or 3 in each direction [19].
  • Non-SCF Mode Execution: Configure the calculation to be a single-shot or non-SCF calculation. This instructs the code to read the converged electron density and Hamiltonian from the previous step and compute the band energies and DOS on the new, finer k-grid without attempting to re-update the density. This is often activated by keywords like SCF.Converge.No or by setting a very high SCF.DM.Tolerance [25].

The Scientist's Toolkit: Essential Computational Reagents

Table 1: Key software and parameters for the two-stage DOS protocol.

Tool/Parameter	Function & Purpose	Example Codes
SCF Algorithms	Solves the Kohn-Sham equations iteratively. DIIS/Pulay is efficient for most systems; GDM is a robust fallback [24].	Q-Chem, SIESTA, ORCA
Mixing Schemes	Accelerates SCF convergence by extrapolating the Hamiltonian or density matrix. Pulay is the default in many codes [25].	SIESTA, VASP
Restart Capability	Allows a new calculation to begin from the final state of a previous one, which is essential for the two-stage protocol [23].	BAND, Quantum ESPRESSO, VASP
K-Point Generator	Automates the generation of uniform k-point meshes in the Brillouin zone.	All major periodic DFT codes
DOS/Band Structure Post-Processing	Calculates the Density of States and band structure from a fixed Hamiltonian, often with advanced interpolation [19].	BAND [23], VASP, Quantum ESPRESSO

Data Presentation and Parameter Optimization

SCF Convergence Tolerances

Achieving a solidly converged SCF in Stage 1 is critical. The following table summarizes typical convergence criteria for different levels of precision, as seen in the ORCA code [26]. Tight convergence is recommended for production calculations, particularly for systems with complex electronic structures.

Table 2: Standard SCF convergence tolerances (as implemented in ORCA) [26].

Criterion	Loose	Medium	Strong	Tight	Description
TolE	1e-5	1e-6	3e-7	1e-8	Energy change between cycles.
TolRMSP	1e-4	1e-6	1e-7	5e-9	RMS density matrix change.
TolMaxP	1e-3	1e-5	3e-6	1e-7	Maximum density matrix change.
TolErr	5e-4	1e-5	3e-6	5e-7	DIIS error norm.

K-Point Sampling Guidelines

The required k-point density is highly system-dependent. The table below offers general guidance, but a convergence test for the DOS itself is the only definitive method.

Table 3: Recommended k-point sampling strategies for different system types.

System Type	Stage 1: SCF Mesh	Stage 2: DOS Mesh	Rationale
Insulator / Molecule	Γ-point or 2x2x2	4x4x4 to 6x6x6	Coarse sampling often sufficient for SCF; finer grid smoothens DOS.
Semiconductor	4x4x4 to 8x8x8	12x12x12 to 20x20x20	Needed to accurately capture the band gap and its edges.
Metal	12x12x12 or finer	24x24x24 or much finer	Essential to resolve sharp features at the Fermi level [19].
1D or 2D Material	Asymmetric mesh (e.g., 12x12x1)	Very dense in periodic directions (e.g., 36x36x1)	Non-periodic directions require minimal sampling (often 1 k-point).

Validation and Technical Considerations

• Convergence Validation: A robust SCF convergence is not just about reaching the default iteration limit. It requires verifying that key metrics—total energy, density matrix, and orbital gradients—have plateaued below the defined thresholds [26]. For production calculations, using Tight settings (e.g., TolE of 1e-8 a.u.) is advisable [26].
• Alternative Strategy - Supercell Scaling: If a code has a hard limit on k-point grid dimensions, one workaround is to build a larger supercell. Using the same k-point sampling (e.g., Γ-point) for a supercell twice the size of the primitive cell is equivalent to sampling the primitive cell with a 2x2x2 mesh, effectively achieving finer sampling for the original structure [28]. However, this complicates the interpretation of band structures due to band folding [28].
• Workflow Automation: Frameworks like atomate2 are designed to automate multi-step computational workflows, including the SCF-converge-then-refine pattern, which is particularly valuable for high-throughput materials screening [29].

Syntax for VASP, CP2K, and BAND

Within computational materials science and drug development research, the accurate calculation of electronic density of states (DOS) is fundamental for understanding material properties, catalytic activity, and molecular interactions. A critical, yet sometimes overlooked, methodological step is the use of a finer k-point mesh during the non-self-consistent field (non-SCF) post-processing step that follows a converged self-consistent field (SCF) calculation. This protocol enhances the resolution of the DOS without the prohibitive computational cost of performing the entire SCF cycle on an ultra-fine k-grid. This document provides detailed application notes and structured protocols for implementing this specific methodology within three prominent electronic structure codes: VASP, CP2K, and BAND. The content is framed within a broader thesis investigating the effects of systematic k-sampling refinement on the accuracy and feature resolution of electronic densities of states.

Comparative Parameter Tables for DOS Calculations

The following tables summarize the essential parameters for DOS calculations across the three software packages, providing a quick reference for researchers.

Table 1: Core Input Parameters for SCF and Non-SCF DOS Calculations

Parameter	VASP	CP2K (Quickstep)	BAND
SCF K-grid	KPOINTS file (automatic mesh) [8]	KPOINTS section (see Table 2)	KSPACING or KPOINTS
Non-SCF K-grid	KPOINTS file (denser mesh)	KPOINTS section (denser mesh)	DOS block (DeltaE, Energies) [30]
Charge Read	ICHARG = 11 (read CHGCAR) [31]	SCF_GUESS RESTART [32]	Default with restart
DOS Flag	LORBIT = 11 (projected DOS) [31]	&PDOS [32]	CalcDOS Yes & CalcPDOS Yes [30]
Smearing	ISMEAR = -5 (tetrahedron) [31]	SMEAR keyword	IntegrateDeltaE Yes [30]

Table 2: K-Point Sampling Syntax Comparison

Software	Sampling Type	Input Syntax Example
VASP [8] [31]	Automatic Γ-centered	```

Automatic mesh 0 Gamma 12 12 12 0 0 0 | | | Band Structure (Line-mode) | Band structure 40 Line Reciprocal 0 0 0 ! Gamma 0.5 0.5 0.0 ! X | | CP2K | K-points subsection (Γ-point) | &KPOINTS FULL_GRID .FALSE. SCHEME MONKHORST-PACK 2 2 2 &END KPOINTS | | BAND [30] | DOS-Specific | DOS DeltaE 0.005 Min -0.35 Max 1.05 File plotfile End ``` |

Unified Workflow for High-Resolution DOS

The established best-practice protocol for obtaining a high-resolution Density of States involves a two-step workflow, which is conceptually consistent across VASP, CP2K, and BAND. The following diagram illustrates this generalized procedure.

Diagram 1: The two-step workflow for high-resolution DOS calculations. The converged charge density from an SCF calculation with a standard k-grid is used as a fixed input for a subsequent non-SCF calculation with a significantly finer k-grid, which produces the final DOS [2] [31].

Software-Specific Protocols

Protocol for VASP

The Vienna Ab initio Simulation Package (VASP) is a widely used code for electronic structure calculations and quantum-mechanical molecular dynamics.

Step 1: SCF Calculation with Standard K-Grid

Objective: Obtain a converged charge density (CHGCAR) and wavefunction (WAVECAR) using a computationally efficient k-point mesh.

• INCAR Parameters:

  For metals, use ISMEAR = -5 (tetrahedron method) for the final DOS calculation [31].

• KPOINTS File: Use a moderately dense, automatically generated mesh. The number of subdivisions should be inversely proportional to the lattice constants [8].

• Execution: Run VASP. Ensure the calculation converges and generates CHGCAR and WAVECAR.

Step 2: Non-SCF Calculation with Fine K-Grid

Objective: Perform a single-shot energy evaluation on a much denser k-point mesh using the precomputed charge density.

• INCAR Parameters:

  Setting ICHARG=11 is the key directive that forces a non-SCF run with a fixed charge density [31].

• KPOINTS File: Use a significantly denser mesh than in Step 1.

• Execution: Run VASP in the same directory, ensuring CHGCAR and WAVECAR from Step 1 are present. The output vasprun.xml file will contain the high-resolution DOS data.

Protocol for CP2K

CP2K performs atomistic simulations using a mixed Gaussian and plane-wave approach, and its k-point sampling syntax differs from VASP.

Step 1: SCF Calculation with Standard K-Grid

Objective: Achieve a converged electron density and obtain a restart file.

• Input File Section (input.inp):

  The SCF_GUESS can be set to RESTART for subsequent calculations [32].

Step 2: PDOS Calculation with Fine K-Grid and Smearing

Objective: Compute the PDOS using a denser k-grid and the pre-converged electron density.

• Input File Section (pdos.inp):

  The combination of SCF_GUESS RESTART and MAX_SCF 1 ensures a non-SCF calculation. The generated PDOS files (e.g., projected_density_of_states) require post-processing convolution with a Gaussian or similar function to produce a smooth DOS plot, as detailed in CP2K exercises [32].

Protocol for BAND

The BAND code, part of the Amsterdam Modeling Suite, specializes in periodic DFT calculations with numerical atomic orbitals.

Step 1: SCF Calculation with Standard K-Grid

Objective: Perform a standard SCF calculation to obtain a converged electron density and a restart file.

• Input File Snippet: The k-grid is typically defined in the main BAND input block or via KSPACING.

Step 2: DOS Calculation with Fine Energy Grid

Objective: Execute a DOS calculation that refines the energy sampling, leveraging the SCF results.

• DOS Block: The DOS block in BAND allows for precise control over the energy range and resolution of the DOS output.

  The DeltaE parameter is crucial; a smaller value results in a finer and smoother DOS [30]. The IntegrateDeltaE keyword controls the algorithm, with the default integrated approach being more robust against missing features [30].

• Gross Populations for pDOS: To obtain atom- or orbital-projected DOS, use the GrossPopulations block.

The Scientist's Toolkit: Essential Research Reagents

This section lists key "research reagents" – the critical input files, parameters, and scripts – necessary for successfully executing the DOS protocols described.

Table 3: Essential Research Reagents for DOS Calculations

Reagent/Files	Function	Considerations
Converged CHGCAR (VASP)	Fixed electron density for non-SCF step [31].	File size can be large for big systems.
WAVECAR / Restart File	Initial guess for wavefunctions in non-SCF run.	Essential for maintaining state.
Fine KPOINTS File	Defines high-sampling Brillouin Zone mesh.	Density should be 2-3x the SCF mesh.
Pseudopotential (POTCAR)	Defines electron-ion interactions.	Consistent for all calculations in a series [33].
Post-Processing Script	Convolutes raw data (e.g., CP2K PDOS) [32].	Choice of smearing (σ) affects line-shape.

Troubleshooting and Optimization Notes

• Missing DOS Features: If the DOS plot lacks expected features or appears too smooth, the most common cause is insufficient k-space sampling. The solution is to restart the DOS calculation with a denser k-grid [30].
• Convergence Warnings in Non-SCF: Non-SCF calculations should converge in 1-2 steps. Failure to do so may indicate problems with the initial SCF CHGCAR/WAVECAR.
• Computational Cost: The memory and time requirements of the non-SCF step scale with the number of k-points and the number of bands. For large systems, it is advisable to test the protocol with a moderately dense k-grid before proceeding to very fine meshes.
• Band Structure Calculations: A related methodology applies to band structure calculations, where the SCF charge density is used in a subsequent non-SCF calculation along a high-symmetry path in the Brillouin zone, as defined by a Line-mode KPOINTS file in VASP [31] or similar input in other codes.

The pursuit of accurate electronic structure calculations, particularly for properties like Density of States (DOS), demands increasingly sophisticated computational methodologies. DOS calculations require high precision in k-space sampling to capture the nuanced electronic behavior essential for understanding material properties relevant to diverse applications, including drug development and catalyst design. Post-SCF (Self-Consistent Field) calculations with finer k-meshes are computationally intensive and methodologically complex, creating significant bottlenecks in research pipelines. Manual execution of these workflows introduces opportunities for error and limits reproducibility.

Atomate2 addresses these challenges directly as a modern, Python-based workflow management system designed specifically for computational materials science. It builds upon the foundation of its predecessor, Atomate, which was created to provide "pre-built workflows to effortlessly compute and analyze various material properties" [34]. This automation framework enables researchers to construct, execute, and manage complex computational protocols with enhanced reliability and efficiency, making sophisticated calculations more accessible across the research community.

Atomate2 Technical Framework and Architecture

Atomate2 implements a modular architecture that separates workflow definition from execution logic, providing flexibility while maintaining rigorous standards for computational reproducibility. The system leverages Python's expressive syntax to create human-readable workflow definitions that can be version-controlled and shared across research groups.

Core Technical Components

The atomate2 framework integrates several specialized components that work in concert to automate materials computations:

• Workflow Manager: Orchestrates the execution of computational tasks, handling dependencies between successive calculations and managing error recovery protocols
• Job Control System: Interfaces with high-performance computing (HPC) resource managers (SLURM, PBS, etc.) to efficiently utilize available computational resources
• Data Parsing Engine: Automatically extracts structured data from output files (e.g., VASP output) and stores it in searchable databases
• Error Handling Module: Implements sophisticated recovery mechanisms for common computational failures, such as convergence issues or node failures

This technical foundation enables researchers to focus on scientific questions rather than computational logistics, significantly accelerating the research lifecycle from hypothesis to results.

K-Point Sampling Strategies for Accurate DOS Calculations

Density of States calculations are particularly sensitive to k-point sampling density due to the need for precise integration over the Brillouin zone. While SCF calculations often converge with moderate k-meshes, accurate DOS requires significantly finer sampling to resolve sharp features in the electronic structure, especially near the Fermi level where electronic transitions occur.

K-Point Convergence Methodology

A systematic approach to k-point convergence is essential for reliable DOS results:

• Initial SCF Calculation: Perform a standard SCF calculation with a moderate k-point mesh to establish the ground-state electron density
• K-Point Convergence Testing: Execute a series of non-SCF calculations with progressively denser k-meshes while monitoring key DOS features
• Convergence Validation: Quantify convergence by tracking integrated DOS values, band edges, and peak positions until changes fall below acceptable thresholds
• Production DOS Calculation: Execute the final DOS calculation with the validated dense k-mesh

The following table summarizes quantitative relationships between k-point density and DOS convergence for representative material systems:

Table 1: K-point Sampling Guidelines for DOS Convergence in Various Material Classes

Material Type	Initial SCF K-points	Recommended DOS K-points	Convergence Threshold (meV)	Typical System Size
Simple Metals	11×11×11	21×21×21	5	10-20 atoms
Semiconductors	13×13×13	25×25×25	2	10-30 atoms
Transition Metal Oxides	15×15×15	31×31×31	1	20-50 atoms
Complex Ceramics	17×17×17	35×35×35	1	40-100 atoms
2D Materials	19×19×1	37×37×1	3	10-30 atoms

K-point Optimization Protocol

The implementation of k-point optimization within atomate2 follows a structured protocol:

K-point Convergence Workflow

Application Note: Automated DOS Workflow with Atomate2

Protocol Implementation

The following application note details a complete workflow for high-fidelity DOS calculations using atomate2, specifically designed for post-SCF calculations with optimized k-point sampling.

Protocol 1: Automated DOS with K-point Convergence

• Workflow Initialization

  • Define initial material structure using pymatgen Structure object
  • Set calculation parameters: functional (PBE, HSE06), energy cutoff, convergence criteria
  • Configure k-point generation strategy (automatic or manual)

• SCF Calculation Setup

  • Generate initial k-point mesh using Monkhorst-Pack scheme
  • Configure INCAR parameters for electronic minimization (ALGO = Normal, PREC = Accurate)
  • Set NEDOS = 1001 for preliminary DOS output

• K-point Convergence Loop

  • Extract eigenvalues from previous calculation
  • Calculate DOS convergence metric (change in integrated DOS)
  • If not converged, generate new k-point mesh with 1.5× density
  • Repeat until convergence threshold is met (typically <2 meV)

• Final DOS Calculation

  • Execute non-SCF calculation with converged k-point mesh
  • Set LORBIT = 11 for projected DOS output
  • Use tetrahedron method (ISMEAR = -5) for k-space integration

• Data Analysis and Storage

  • Parse DOS data to database
  • Calculate integrated DOS and band centers
  • Generate visualization outputs

Workflow Visualization

The complete automated workflow for DOS calculation with k-point convergence is depicted below:

Automated DOS Calculation Workflow

Research Reagent Solutions: Computational Tools for Materials Research

The computational materials science ecosystem comprises specialized software tools that function as "research reagents" - essential components that enable specific research capabilities. The following table catalogs key solutions relevant to automated workflow management and electronic structure calculations.

Table 2: Essential Computational Tools for Automated Materials Research

Tool Name	Category	Primary Function	Integration with Atomate2
VASP	Electronic Structure	DFT calculations for materials properties	Direct (primary calculator)
VASPKIT	Pre/Post-processing	Utilities for VASP input/output processing	Complementary [35]
Pymatgen	Materials Analysis	Python library for materials analysis	Core dependency [34]
FireWorks	Workflow Management	Workflow scheduling and job management	Foundation for atomate
Custodian	Error Handling	Job management and error recovery	Core dependency [34]
Matgl	Machine Learning	Graph neural networks for materials	Emerging integration

Advanced Protocols for Specialized Applications

Protocol 2: Hybrid Functional DOS Calculations

Hybrid functionals (e.g., HSE06) provide improved electronic structure accuracy but require substantial computational resources. This protocol extends the basic DOS workflow for hybrid calculations:

• Initial SCF with Standard Functional

  • Perform PBE calculation with standard k-point mesh
  • Converge charge density to 10⁻⁶ eV

• Wavefunction Conversion

  • Transform Kohn-Sham orbitals to hybrid functional basis
  • Preserve wavefunctions for subsequent calculations

• Hybrid Functional SCF

  • Execute HSE06 calculation using pre-converged wavefunctions
  • Use reduced k-point mesh initially (50-70% of PBE mesh)

• DOS with Exact Exchange

  • Perform non-SCF hybrid functional DOS with fine k-mesh
  • Account for exact exchange in final electronic structure

Protocol 3: Projected DOS for Complex Systems

Projected DOS (PDOS) analysis decomposes the electronic structure by atomic species, orbitals, or specific atomic sites:

• Wavefunction Generation

  • Perform converged SCF calculation with high-precision settings
  • Set LREAL = .FALSE. for accurate projection

• Projection Setup

  • Configure projection parameters (LORBIT = 11)
  • Define atomic groups for specific PDOS analysis

• Projected DOS Calculation

  • Execute single-shot calculation with fine k-mesh
  • Generate site-projected and orbital-projected DOS

• Data Integration

  • Sum projections across selected atomic groups
  • Calculate orbital-resolved band centers
  • Generate visualization of orbital contributions

Data Management and Analysis Framework

Atomate2 implements a comprehensive data management strategy that ensures computational results are structured, searchable, and reusable. The system automatically parses output files and stores key results in document databases with rich metadata, including:

• Calculation parameters (functional, k-points, convergence settings)
• System information (structure, composition, symmetry)
• Performance metrics (computational time, memory usage)
• Scientific results (total energy, band gap, DOS, forces)

This structured approach to data management facilitates high-throughput screening and machine learning applications by providing consistent, well-documented computational results.

Atomate2 represents a significant advancement in workflow management for computational materials science, particularly for precision tasks like DOS calculations with fine k-sampling. By automating the complex sequence of calculations, error handling, and data management, the system enables researchers to focus on scientific interpretation rather than computational details. The protocols outlined in this application note provide a foundation for implementing these methodologies in diverse research contexts, from fundamental materials discovery to applied drug development projects.

As computational materials science continues to evolve, integration with machine learning approaches and multi-fidelity computational strategies will further enhance the capabilities of workflow managers like atomate2. These advancements promise to accelerate materials discovery and optimization, ultimately contributing to faster development cycles in fields ranging from pharmaceuticals to energy technologies.

In the computational analysis of crystalline materials, the calculation of electronic properties like the band structure and density of states (DOS) is fundamental. For two-dimensional materials like graphene, the choice of k-point sampling is particularly critical, as it directly influences the accuracy of the predicted electronic structure. This application note details a robust methodology for calculating the band structure and DOS of graphene, emphasizing the use of a dense k-path during the non-self-consistent field (NSCF) calculation step to achieve high-fidelity results. This approach aligns with a broader research thesis that systematic convergence testing, especially using finer k-sampling for post-processing, is essential for obtaining reliable electronic structure data [15].

The workflow typically involves two primary stages: a self-consistent field (SCF) calculation to determine the ground-state electron density, followed by an NSCF calculation on a more refined k-point grid or path to compute the band structure and DOS [15]. For graphene, whose unique Dirac cone dispersion is sensitive to sampling, a dense k-path is indispensable for accurately capturing its electronic characteristics.

Computational Workflow and Methodology

The general protocol for obtaining the band structure and DOS involves a sequential process of self-consistent and non-self-consistent calculations, leveraging different k-point sampling strategies at each stage. The figure below illustrates the complete workflow.

As shown in the workflow diagram, the process begins with a single Self-Consistent Field (SCF) calculation to obtain the converged charge density [15]. This initial step uses a relatively coarse, uniform k-point grid (e.g., a Γ-centred Monkhorst-Pack grid like 12 12 1) [15]. The resulting charge density is then used in two separate Non-Self-Consistent Field (NSCF) calculations: one employing a dense uniform k-grid for an accurate DOS and another using a densely spaced path of high-symmetry points in the Brillouin zone for the band structure [15]. This separation is computationally efficient, as the ground state needs to be calculated only once.

Key Research Reagent Solutions

The following table details the essential computational "reagents" and their functions in this protocol.

Table 1: Essential Computational Tools and Parameters for Band Structure and DOS Calculations

Item	Function/Description	Example/Value
Pseudopotential	Represents the interaction between valence electrons and the ion core [15].	C.pbe-rrkjus.UPF (PBE ultrasoft for Carbon) [15]
k-Point Grid (SCF)	Samples the Brillouin zone for the initial ground-state calculation [15].	12 12 1 (Monkhorst-Pack grid for 2D systems) [15]
k-Path (NSCF)	Defines the high-symmetry path for the band structure calculation [15].	Γ (0,0,0) → M (0.5,0,0) → K (1/3,1/3,0) → Γ (0,0,0) [15]
Broadening Method	Smears discrete electronic levels into a continuous DOS [36].	Gaussian (0.05–0.2 eV) or Tetrahedron method [36] [37]
Exchange-Correlation Functional	Approximates the quantum mechanical exchange-correlation energy [15].	PBE (Perdew-Burke-Ernzerhof) [15]

Detailed Experimental Protocol

Self-Consistent Field (SCF) Calculation

The first step is to perform an SCF calculation to obtain the converged charge density and wavefunctions.

• System Definition: Specify the crystal structure. For graphene, this involves defining a hexagonal cell with two carbon atoms.
  • Lattice Constant: celldm(1) = 4.6530 (a.u.), corresponding to 2.462 Å [15].
  • Atomic Positions: in crystal coordinates: C 0.666666666 0.333333333 0.000000000 C 0.333333333 0.666666666 0.000000000 [15].
• K-Point Sampling: Use a Γ-centred Monkhorst-Pack grid. A grid of 12 12 1 is a typical starting point for good convergence [15].

• Execution: Run the SCF calculation. In Quantum ESPRESSO, this is done with the pw.x code [15]:

• Post-Processing:

  • Extract the Fermi energy from the output file for later use [15]:

  • Save the output directory (e.g., mol.save/) containing the charge density for the subsequent NSCF calculations [15].

Density of States (DOS) Calculation

The DOS is calculated using a fixed charge density from the SCF run but with a much denser k-point grid in a separate NSCF calculation.

• NSCF Input Preparation: Create an input file (nscf.in) similar to scf.in but with key modifications [15]:
  • Set calculation = 'nscf'.
  • Increase the k-point grid significantly. For a high-quality DOS, a 40 40 1 grid or denser is recommended [15].
  • Specify the path to the previously saved charge density (prefix = './mol').

• Run NSCF Calculation:

• Compute DOS: Use a dedicated DOS calculation tool (e.g., dos.x in Quantum ESPRESSO) that interpolates the eigenvalues onto a smooth energy spectrum [15]. The input file (dos.in) should specify the energy range and broadening.

• Shift and Plot: Shift the energy axis so that the Fermi level is at zero, and plot the DOS using a tool like gnuplot [15] [37].

Band Structure Calculation with a Dense k-Path

The band structure requires an NSCF calculation where k-points are defined along high-symmetry lines in the Brillouin zone.

• Define the High-Symmetry Path: The path must traverse critical points. For graphene's hexagonal Brillouin zone, the standard path is Γ → M → K → Γ [15]. The coordinates of these points are:
  • Γ: (0, 0, 0)
  • M: (0.5, 0, 0)
  • K: (1/3, 1/3, 0) These coordinates are in units of the reciprocal lattice vectors. [15]
• NSCF Input for Bands: Create a bands.in input file [15]:
  • Set calculation = 'bands'.
  • Use the charge density from the SCF run (prefix = './scf-mol').
  • Replace the automatic k-point grid with a K_POINTS block that lists the high-symmetry path. A large number of points (e.g., 100) between special points ensures a smooth band structure [15].

• Run Band Structure Calculation:

• Data Extraction and Plotting: The output file contains eigenvalues for each k-point. Use a script (e.g., Qe_Bands.sh [15] or cp2k_bs2csv.py [38]) to parse the data into a plottable format (k-point index vs. energy). Plot the bands, setting the Fermi energy to zero and marking the high-symmetry points.

Data Presentation and k-Sampling Convergence

The choice of k-sampling in the NSCF step profoundly impacts the quality of the results. The following diagram illustrates the logical relationship between k-sampling strategy and the resulting electronic structure output.

Quantitative Impact of k-Grid Density

Systematically increasing the k-point grid density in the NSCF calculation is crucial for converging the DOS. The table below summarizes the typical effects of varying the k-grid.

Table 2: Convergence of DOS and Computational Cost with k-Grid Sampling

k-Grid (SCF)	k-Grid (NSCF for DOS)	Key Observations on DOS	Relative Computational Cost
12 12 1	20 20 1	Features are recognizable but may be jagged; Dirac point not perfectly resolved.	Low
12 12 1	30 30 1	Smoother DOS; linear behavior near Dirac point becomes clearer.	Medium
12 12 1	40 40 1	Well-converged, smooth DOS; linear dependence near EF is accurately captured [15].	High

For the band structure, the density of the k-path (the number of points between high-symmetry points) determines the smoothness of the bands. A sparse path will produce a jagged, discontinuous plot, while a dense path (e.g., 100 points between Γ and M) results in a smooth, publication-quality figure [15].

Discussion

The protocol outlined above demonstrates that a two-step SCF/NSCF approach is both computationally efficient and physically sound. The core principle is that the ground-state charge density, obtained from an SCF calculation with a moderately dense k-grid, is a robust quantity. Subsequent analysis of derived electronic properties can then be performed with different, often much denser, k-point samplings without repeating the expensive SCF cycle.

This methodology directly supports the broader thesis that finer k-sampling in post-SCF calculations is a critical research practice. For graphene, this is evident in the precise resolution of the linear Dirac cone and the van Hove singularities. In more complex materials, such as those with flat bands or topological states, dense k-sampling can be the difference between correctly identifying a material's character and missing it entirely. Furthermore, advanced analysis like projected DOS (PDOS) and fatband structures, which assign orbital or atomic character to bands and DOS, also rely fundamentally on well-converged k-sampled wavefunctions [37] [39].

In conclusion, adopting a systematic protocol that emphasizes dense k-path sampling for NSCF calculations is indispensable for achieving accurate and reliable electronic structure results, forming a cornerstone of rigorous computational materials science research.

Overcoming Common Hurdles: Ensuring Convergence and Managing Resources

Diagnosing and Fixing SCF Convergence Issues with Dense Meshes

In the pursuit of accurate electronic properties such as Density of States (DOS) within first-principles calculations, employing a dense k-point mesh is often essential for achieving well-converged results. However, this practice frequently introduces significant challenges for the Self-Consistent Field (SCF) convergence procedure. A dense mesh increases the complexity of the electronic structure landscape, often leading to oscillations, stagnation, or failure of the SCF cycle. This application note provides a structured methodology for diagnosing the root causes of these failures and implementing robust solutions, with a specific focus on workflows where a converged SCF is a prerequisite for subsequent high-quality DOS analysis. The protocols herein are designed to integrate seamlessly into a research pipeline aimed at using finer k-sampling for post-SCF DOS calculation research.

Diagnosing the Problem: A Systematic Approach

Before applying corrective measures, a systematic diagnosis is crucial. The symptoms of SCF failure can often point toward their underlying cause. The following table categorizes common manifestations and their likely origins, particularly in the context of calculations involving dense k-meshes.

Table 1: Common SCF Failure Symptoms and Their Diagnostic Significance

Symptom	Description	Likely Causes
Charge/Spatial Oscillations [40]	The total energy or density matrix oscillates between values without settling.	Inadequate damping; DIIS extrapolation becoming unstable; near-degenerate states.
Convergence Stagnation [40]	The energy or error converges to a point, then stops improving.	Insufficient SCF iterations (MaxIter); numerical noise; shallow energy landscape.
Divergence [40]	The SCF energy increases drastically or becomes non-physical.	Poor initial guess; unreasonable initial geometry; large basis set superposition error.
TRAH Slow Convergence [40]	The robust but expensive Trust Radius Augmented Hessian (TRAH) algorithm is activated but is very slow.	Extremally difficult convergence; may require adjustment of AutoTRAH parameters or a better initial guess.

The diagnostic workflow below provides a logical pathway to identify the problem based on observed behavior.

Solution Strategies and Experimental Protocols

Once a diagnosis is established, a targeted solution can be applied. The strategies below are ordered from simple, general fixes to more advanced, system-specific protocols.

Core Solution Strategies

Table 2: Summary of Key SCF Convergence Solutions

Strategy	Mechanism	Typical Command/Keyword	Applicability
Improve Initial Guess	Starts SCF closer to the solution, reducing initial large fluctuations.	! MORead (ORCA), guess=read (Gaussian), USPM (VASP) [40] [41]	Universal first step, especially after geometry or basis set changes.
Algorithm Switching	Replaces default DIIS with more robust, quadratically convergent or direct minimization methods.	SCF=QC or SCF=XQC (Gaussian) [42] [43], SCF_ALGORITHM=GDM (Q-Chem) [44]	For oscillating or stagnant convergence where DIIS fails.
Damping & Level Shifting	Damping mixes old and new densities; level shifting increases HOMO-LUMO gap to prevent orbital mixing.	! SlowConv (ORCA) [40], SCF=vshift=400 (Gaussian) [41]	Systems with small band gaps (e.g., metals, open-shell TMs).
DIIS Enhancement	Increases the history of Fock matrices for better extrapolation, resets numerical noise.	DIISMaxEq 20 (ORCA) [40], directresetfreq 5 (ORCA) [40]	For systems where DIIS is unstable or slowly converging.
Convergence Criterion Relaxation	Relaxes the convergence threshold for the initial SCF.	SCF_CONVERGENCE 6 (Q-Chem) [44], SCF=conver=6 (Gaussian) [41]	For initial/pre-computation only. Not for final production runs.

Detailed Protocol for DOS Workflow with Dense k-Meshes

This protocol outlines a robust, multi-stage workflow designed to obtain a converged wavefunction for a dense k-mesh DOS calculation.

Objective: To efficiently achieve a fully converged SCF ground state using a dense k-point mesh, from which a non-SCF (NSCF) DOS calculation can be performed.

Principle: Decouple the challenging SCF convergence from the dense k-point DOS calculation. First, converge the SCF on a computationally cheaper setup, then use the resulting wavefunction as an expert guess for a single-shot NSCF calculation on the dense mesh [23].

Materials (The Scientist's Toolkit):

Table 3: Essential Computational Reagents for SCF Convergence

Reagent / Software Feature	Function	Example Implementations
Pseudopotential (PP)/Basis Set	Defines the electron-ion interaction and wavefunction expansion. Choosing the right balance is critical.	NCPPs, USPPs [45]
Coarse k-point Mesh	Reduces the number of k-points for the initial SCF, drastically speeding up convergence.	KSPACING, KPOINTS (VASP)
Wavefunction File	Serves as the high-quality initial guess for subsequent calculations.	gbw (ORCA), WAVECAR (VASP), checkpoint file (Gaussian)
Damping Factor	Stabilizes the SCF by mixing a fraction of the previous density matrix with the new one.	$scfdamp (TURBOMOLE) [46]
Hybrid SCF Algorithm	Combines the speed of DIIS initially with the robustness of fallback algorithms like GDM.	SCF_ALGORITHM=DIIS_GDM (Q-Chem) [44]

Step-by-Step Procedure:

• Geometry Optimization: Perform a full geometry optimization using a moderate k-point mesh and standard SCF settings. Ensure the geometry is fully converged. A reasonable geometry is a prerequisite for SCF convergence [47].
• Initial SCF on Coarse Mesh:
  • Using the optimized geometry, perform a single-point energy calculation.
  • Use the same moderate k-point mesh from the geometry optimization.
  • Employ standard SCF convergence algorithms (e.g., DIIS). If convergence fails, apply basic fixes like increasing MaxIter or using SlowConv [40].
  • Output: A fully converged wavefunction file (e.g., WAVECAR, gbw).
• Dense Mesh DOS Calculation (NSCF):
  • Construct a new input file for the DOS calculation.
  • Set the k-point mesh to the desired dense sampling for the DOS.
  • Critical: Read the converged wavefunction from Step 2 as the initial guess (guess=read, MORead, WAVECAR).
  • Run this as a non-self-consistent calculation if possible. This means the code will take the initial wavefunction and diagonalize it once on the new k-mesh to compute the DOS, bypassing the SCF cycle entirely. If an NSCF mode is not available, run a standard SCF job but with the expert guess; it should converge in very few cycles.

Troubleshooting Advanced Cases:

For pathologically difficult systems (e.g., open-shell transition metal complexes, magnetic materials, or large clusters), the standard protocol may fail. In such cases, implement these advanced strategies within Step 2:

• Protocol 3.2.A: For Severe Oscillations. Use the SCF=QC (Gaussian) [42] [43] or SCF_ALGORITHM=GDM (Q-Chem) [44] keywords. These algorithms are more robust but significantly more computationally expensive per iteration.
• Protocol 3.2.B: For Systems with Small HOMO-LUMO Gaps. Apply level shifting to virtual orbitals. For example, in Gaussian, use SCF=vshift=400 to shift orbital energies by 0.4 Hartree, which artificially increases the HOMO-LUMO gap and prevents excessive mixing between occupied and virtual orbitals [41].
• Protocol 3.2.C: For Large Metallic Systems. Enable Fermi smearing (e.g., SCF=Fermi in Gaussian) [42] or similar occupational smearing. This allows fractional occupation of orbitals around the Fermi level, which can help resolve convergence issues in metals.

Successfully converging the SCF for dense k-meshes is a common hurdle in computational research. By adopting a systematic diagnostic approach and implementing a staged workflow that separates the SCF convergence from the final high-k-point calculation, researchers can significantly improve reliability and efficiency. The strategies and protocols detailed in this note provide a clear pathway to overcome these challenges, ensuring that robust electronic structure results, such as high-resolution DOS, can be obtained consistently. This methodology is particularly vital for thesis research and high-throughput materials screening where reproducibility and accuracy are paramount.

In the context of research focused on employing finer k-sampling for Density of States (DOS) post-self-consistent field (post-SCF) calculations, managing computational cost is a fundamental challenge. First-principles methods, particularly those based on Density Functional Theory (DFT), provide a powerful framework for investigating electronic structures but can become prohibitively expensive for large systems or when high numerical accuracy is required [48] [49]. The computational expense scales significantly with system size, the density of the k-point grid used for Brillouin zone sampling, and the choice of the basis set [50] [8]. This article outlines practical strategies and detailed protocols to manage these costs effectively without compromising the reliability of the results, with a specific emphasis on DOS calculations within a broader thesis research framework.

Background and Core Concepts

The SCF Cycle and Post-SCF Analysis

Density Functional Theory (DFT) has emerged as a leading first-principles approach in quantum mechanical investigations due to its favorable balance between computational expense and chemical accuracy [48]. The Kohn-Sham formalism of DFT expresses the ground state electronic energy as a functional of the electron density, which is constructed from single-electron Kohn-Sham orbitals [48]. The process of finding a self-consistent solution for these orbitals is known as the self-consistent field (SCF) cycle. Once the SCF cycle converges and the ground-state electron density is obtained, post-SCF analyses, such as band structure and Density of States (DOS) calculations, are performed. These analyses provide crucial information about electronic properties but require a denser sampling of k-points in the Brillouin zone than what is typically used for the initial SCF calculation to achieve sufficient energy resolution [36] [8].

Key Computational Bottlenecks

The primary computational bottlenecks in DFT calculations for large systems and dense grids include:

• System Size Scaling: The computational cost of traditional ab initio methods scales poorly with system size, which limits their application to large or complex systems [51].
• K-Point Sampling: Achieving a converged DOS often requires a very fine k-point mesh. The number of k-points is approximately inversely proportional to the unit cell size; larger supercells require fewer k-points, but the total number of atoms increases [8].
• Basis Set Choice: The size and type of the atomic basis set (e.g., Plane-Wave vs. Localized Basis) directly impact memory requirements and computation time [52].
• Algorithmic Complexity: The calculation of exact exchange in hybrid functionals or the treatment of strong electron correlation are computationally demanding [53] [50].

Strategic Approaches for Cost Management

K-Point Sampling Strategies

Efficient k-point sampling is critical for balancing accuracy and computational cost.

Table 1: K-Point Generation Schemes

Scheme	Description	Best Use Cases
Γ-Centered Mesh [8]	Generates a uniform mesh centered on the Gamma point.	General-purpose SCF calculations on symmetric systems.
Monkhorst-Pack Mesh [8]	Generates a mesh that may converge faster than Gamma-centered meshes for some systems.	Systems where a slight offset from Gamma point is beneficial; caution with symmetry.
Automatic K-Path [52]	Automatically generates a high-symmetry path through the Brillouin zone.	Band structure and DOS post-SCF calculations along symmetry lines.

A key strategy for DOS calculations is a two-step process: first, perform the SCF calculation with a moderately dense k-point grid to obtain a converged charge density, and then perform a non-SCF (post-SCF) calculation on a much denser k-point grid to compute the DOS with high resolution [36] [50]. This is more efficient than running the entire SCF cycle with an ultra-dense grid. For example, a SCF calculation might use a 4x4x4 mesh, while the subsequent DOS calculation could use a 16x16x16 mesh [36].

Basis Set and Integration Grid Optimization

The choice of basis set and numerical integration grid can be optimized for performance.

Table 2: Basis Set and Integration Options

Method	Function	Impact on Cost
Frozen Core Approximation [52]	Treats core electrons as inert, reducing the number of active electrons.	Significantly reduces computational effort.
Numerical Integration Grid [52]	Defines the grid for numerically integrating functionals (e.g., BeckeGrid).	Lower quality grids speed up calculations but reduce accuracy.
Density Fitting (RI) [50]	Uses an auxiliary basis to approximate two-electron integrals.	Greatly accelerates Hartree-Fock and hybrid DFT calculations.

Using a Double Zeta Polarized (DZP) basis set often offers a good compromise between accuracy and cost for initial calculations [52]. For the numerical integration grid used in computing the exchange-correlation potential, selecting a "Good" or "Normal" quality instead of "Excellent" can provide significant speedups with minimal accuracy loss for many properties [52].

Advanced Algorithms and Parallelization

Leveraging advanced algorithms and efficient parallelization is essential for large-scale calculations.

• Density Fitting: Also known as the Resolution-of-Identity (RI) technique, this method is widely used to speed up the computation of two-electron integrals in Hartree-Fock and hybrid DFT calculations [50]. PySCF supports both plane-wave (FFTDF) and Gaussian (GDF) density fitting for periodic systems [50].
• Multigrid Methods: For pure DFT calculations on large unit cells, multigrid algorithms (e.g., multigrid.multigrid in PySCF) can be more efficient than standard plane-wave approaches [50].
• Smearing: For metallic systems or small-gap semiconductors, applying a smearing function (e.g., Fermi-Dirac) to orbital occupations can improve SCF convergence and reduce the number of k-points needed to reach the thermodynamic limit [50].
• Machine Learning Potentials: Machine learning interatomic potentials (MLIPs) can achieve near-ab initio accuracy at a fraction of the computational cost, making them suitable for large-scale molecular dynamics simulations that would be infeasible with direct DFT [51] [49].

Detailed Experimental Protocols

Protocol 1: High-Throughput DOS Workflow

This protocol is adapted from high-throughput frameworks for lattice dynamics and electronic structure calculations [51] [49].

• Initial Structure Optimization

  • Objective: Obtain a precise, fully relaxed crystal structure.
  • Method: Use a solid-optimized exchange-correlation functional like PBEsol [49].
  • Parameters: A k-point spacing of 0.04 Å⁻¹ or a Gamma-centered 4x4x4 mesh is often sufficient for the initial relaxation [8].

• SCF Calculation

  • Objective: Generate a converged electron density and wavefunction.
  • Parameters: Use a k-point grid denser than the relaxation step (e.g., 6x6x6). Employ a moderate basis set (e.g., TZP). For hybrid functionals, activate density fitting (GDF) to reduce cost [50].

• Post-SCF DOS Calculation

  • Objective: Compute the electronic Density of States with high energy resolution.
  • Method: Perform a non-SCF calculation on a significantly denser k-point grid.
  • Parameters: Use a k-point grid of at least 16x16x16 [36]. The tetrahedron method is recommended for generating the DOS spectrum from the eigenvalues [36].

The workflow for this protocol is summarized in the diagram below.

Protocol 2: DOS Calculation for a Silicon Crystal

This protocol provides a concrete example of calculating the DOS for a silicon crystal in the diamond structure using an LCAO calculator, as implemented in codes like QuantumATK [36].

• Configuration Setup:

• Calculator Configuration:

• DOS Calculation:

• Spectrum Generation and Plotting:

The Scientist's Toolkit: Essential Computational Reagents

Table 3: Key Software and Computational Tools

Tool / "Reagent"	Type	Primary Function in Calculation
VASP [49] [8]	Software Package	A widely used plane-wave DFT code for periodic systems, featuring robust k-point sampling and DOS analysis.
PySCF [50]	Software Package	A Python-based quantum chemistry framework supporting periodic Hartree-Fock and DFT with various integral schemes.
QuantumATK [36]	Software Package	A platform for atomic-scale modeling with advanced analysis tools, including the DensityOfStates class.
FFTDF / GDF [50]	Algorithm (Density Fitting)	Accelerates the computation of electron repulsion integrals in periodic calculations.
Tetrahedron Method [36]	Algorithm (DOS)	Provides a more accurate DOS spectrum from a discrete set of k-points compared to Gaussian broadening.
Setyawan-Curtarolo K-Path [52]	Algorithm (k-path)	Automatically generates high-symmetry paths in the Brillouin zone for band structure and DOS calculations.
Machine Learning Potentials (MLIPs) [51] [49]	Model / Algorithm	Fast, accurate surrogate models for DFT energies and forces, enabling large-scale molecular dynamics.

To select the most appropriate protocol, researchers can follow the decision logic illustrated below.

In conclusion, managing computational costs for DOS calculations with dense k-grids requires a multi-faceted approach. The strategies outlined—including a two-step SCF/post-SCF process, efficient k-point sampling, basis set optimization, and leveraging advanced algorithms like density fitting and machine learning potentials—provide a robust framework for researchers. By applying these protocols and utilizing the provided decision framework, scientists can effectively balance numerical accuracy with computational feasibility, enabling the successful execution of their research objectives within the scope of a broader thesis on high-resolution electronic structure analysis.

Addressing Disk Space Limitations for Large-Scale k-Point Calculations

Density functional theory (DFT) has revolutionized our ability to predict and explore materials properties computationally. A critical aspect of these calculations involves k-point sampling, which discretizes the continuous Brillouin zone to make computations feasible. For properties like density of states (DOS) that require high precision, researchers must employ exceptionally fine k-point meshes in non-self-consistent field (NSCF) calculations run after obtaining a converged charge density. This approach, while essential for accuracy, generates substantial data output that can quickly overwhelm storage resources, particularly in high-throughput computational workflows and for complex systems requiring dense sampling.

The fundamental challenge arises because each additional k-point increases the volume of wavefunction and eigenvalue data written to disk. While the self-consistent field (SCF) calculation establishes the electronic ground state on a practical k-point grid, subsequent NSCF calculations for DOS typically require 5-10 times denser sampling to achieve sufficient energy resolution. This guide addresses the strategic balance between computational accuracy and storage limitations, providing protocols for optimizing k-point calculations within constrained disk environments.

Theoretical Foundation: SCF versus NSCF Calculation Paradigms

The Kohn-Sham Equations and Self-Consistency

At the heart of DFT lies the Kohn-Sham equation, which must be solved self-consistently:

$$ \left(-\frac{\hbar^2}{2m}\nabla^2 + \hat{V}[\rho]\right)\psi{bk} = E{bk}\psi_{bk} $$

Here, $\psi_{bk}$ represents the Kohn-Sham orbital for band $b$ at k-point $k$, and $\hat{V}[\rho]$ is the potential functional of the electron density $\rho(\mathbf{r})$. The self-consistent field method iteratively solves this equation until the input and output densities converge, ensuring a self-consistent solution [6].

Role of NSCF Calculations for DOS and Band Structure

Once convergence is achieved in the SCF calculation, the resulting charge density $\rho(\mathbf{r})$ provides a accurate representation of the electronic ground state. This density can then be used to construct the Kohn-Sham Hamiltonian without further updating $\rho$, enabling non-self-consistent field calculations [54]. NSCF calculations serve critical purposes:

• Band structure visualization: Solving the Kohn-Sham equations along high-symmetry directions in the Brillouin zone
• Density of states calculation: Using a dense uniform k-point mesh to accurately determine the electronic DOS
• Projected DOS analysis: Decomposing the DOS by atomic site, orbital, or element
• Empty state calculations: Determining unoccupied states for spectroscopic simulations

The key advantage of NSCF calculations lies in their ability to use different k-point sets than the original SCF calculation, specifically optimized for the desired property [55].

Disk Space Challenges in Large-Scale k-Point Calculations

Data Volume Scaling with k-Point Density

The relationship between k-point density and data storage requirements follows approximately cubic scaling for three-dimensional systems. As shown in Table 1, increasing k-point sampling from a basic to excellent quality mesh can increase the number of k-points by over 20 times, with corresponding growth in output file sizes.

Table 1: K-point density recommendations and relative computational cost for different quality settings

Quality Setting	Typical k-point Density Range	Relative CPU Time	Energy Error/Atom (eV)
Gamma-Only	1×1×1	1×	3.3 [3]
Basic	3×3×3 to 5×5×5	2×	0.6 [3]
Normal	5×5×5 to 9×9×9	6×	0.03 [3]
Good	9×9×9 to 13×13×13	16×	0.002 [3]
VeryGood	13×13×13 to 17×17×17	35×	0.0001 [3]
Excellent	17×17×17 to 21×21×21	64×	Reference [3]

Critical Storage Bottlenecks

The primary storage challenges in large-scale k-point calculations include:

• Wavefunction storage: The single-particle orbitals $\psi_{bk}$ for each k-point and band constitute the largest memory and storage requirement
• Eigenvalue files: Energy eigenvalues $E_{bk}$ for all k-points and bands
• Charge density grids: The real-space charge density representation
• Projected DOS data: Additional storage for orbital-, site-, or element-projected DOS

For a typical NSCF DOS calculation with 10,000 k-points and 500 bands, wavefunction storage alone can require terabytes of disk space, creating significant infrastructure challenges for research groups.

Strategic Approaches to Storage Optimization

k-Point Optimization and Convergence Testing

Systematic convergence testing helps identify the minimum k-point sampling required for target precision. As demonstrated in recent uncertainty quantification approaches, the relationship between k-point density and error follows predictable asymptotic behavior, enabling optimized parameter selection [56]. The recommended protocol:

• Perform incremental convergence tests starting from coarse k-point meshes
• Focus on the property of interest (formation energy, band gap, DOS resolution) rather than total energy alone
• Leverage error cancellation in energy differences when possible
• Use automated optimization tools like pyiron's implementation for parameter selection [56]

Computational Workflow Design

The separation of SCF and NSCF calculations represents the most significant optimization for storage management. Figure 1 illustrates this optimized workflow:

Figure 1: Optimized SCF-NSCF workflow for efficient k-point calculations

This workflow minimizes redundant calculations by reusing the converged charge density from the SCF step for multiple NSCF calculations with different k-point sets tailored to specific properties.

Experimental Protocols for Storage-Efficient Calculations

Protocol 1: Two-Step DOS Calculation with Quantum ESPRESSO

Table 2: Key research reagents and computational parameters for storage-efficient DOS calculations

Component	Function	Recommended Settings
SCF k-point grid	Initial charge density convergence	4×4×4 to 8×8×8 depending on system size
NSCF k-point grid	DOS calculation	12×12×12 to 24×24×24 for high resolution
degauss parameter	DOS broadening	0.01-0.05 Ry for metals, 0.001-0.01 Ry for semiconductors
verbosity setting	Control output size	"high" for full eigenvalue output, "low" for reduced output
filpdos	Projected DOS output file	Element- or orbital-projected DOS data

Step-by-Step Implementation:

• SCF Calculation Setup

  • Use calculation = "scf" in the Quantum ESPRESSO input file
  • Select a moderate k-point grid sufficient for charge density convergence
  • Example: K_POINTS {automatic} with 6 6 6 0 0 0 for a cubic system

• NSCF Calculation for DOS

  • Use calculation = "nscf" with the same converged charge density
  • Implement a denser k-point grid: 20 20 20 0 0 0 or finer
  • Increase the number of bands if empty states are needed
  • Set diago_full_acc = .true. for accurate diagonalization

• DOS Post-Processing

  • Use dos.x with appropriate broadening (degauss)
  • Specify energy range relevant to your system
  • For projected DOS, use projwfc.x to decompose by atomic site or orbital [55]

Protocol 2: Band Structure Calculations with K-Path Sampling

Implementation Steps:

• Initial SCF Calculation

  • Perform standard SCF with automatic k-point grid
  • Ensure full convergence of total energy and charge density

• Band Structure NSCF Calculation

  • Use calculation = "bands" in Quantum ESPRESSO
  • Define high-symmetry k-path using tools like SeekPath
  • Example k-point input for crystal_b format:

• Band Structure Plotting

  • Use bands.x for post-processing
  • Combine with DOS plots for comprehensive electronic structure analysis [55]

Protocol 3: Storage-Aware Hybrid Functional Calculations

For more accurate band gaps and excited-state properties, hybrid functionals like HSE06 are essential but computationally demanding:

• SCF with Standard Functional

  • Perform initial calculation with PBE or similar GGA functional
  • Achieve charge density convergence

• One-Shot Hybrid Calculation

  • Use converged PBE density as starting point for hybrid functional NSCF
  • Significantly reduces number of expensive hybrid cycles [57]

• Band Structure with Hybrid Functional

  • Perform hybrid functional band structure as NSCF calculation
  • Example: HSE06 with 25% exact exchange for Si, or material-specific tuned exchange [57]

Advanced Techniques for Storage Management

Data Compression and Selective Output

Modern computational packages offer options to reduce storage requirements:

• Binary format selection: Use efficient binary formats instead of ASCII
• Wavefunction restart files: Configure to keep only essential restart information
• Eigenvalue-only output: For DOS calculations, sometimes only eigenvalues are needed
• Subset extraction: Save only relevant energy windows near the Fermi level

Computational Resource Trade-offs

Strategic decisions can balance computational and storage costs:

• k-point parallelism: Distribute k-points across computational nodes to reduce memory per node
• Symmetry reduction: Use crystal symmetry to reduce the number of irreducible k-points
• Incremental refinement: Start with coarse sampling, selectively refine regions of interest
• Sparse storage: For large systems, use sparse matrix formats where applicable

Effective management of disk space limitations for large-scale k-point calculations requires a multifaceted approach combining theoretical understanding, computational optimization, and strategic workflow design. The protocols presented here enable researchers to maximize the scientific return from their computational investments while working within practical storage constraints.

The SCF/NSCF separation strategy remains the most powerful technique, allowing researchers to reuse the computationally expensive converged charge density for multiple specialized calculations with different k-point samplings optimized for specific properties. As computational methods evolve, emerging techniques including automated convergence parameter optimization [56] and advanced data compression will further alleviate storage bottlenecks, enabling ever more accurate electronic structure calculations for materials discovery and design.

For research groups operating in storage-constrained environments, prioritization of calculations and selective data retention policies are essential. By implementing the protocols outlined in this application note, computational researchers can significantly enhance their productivity while maintaining the high accuracy standards required for cutting-edge materials research.

Resolving Discrepancies Between DOS and Band Structure Data

Within the framework of density functional theory (DFT) simulations, a common challenge arises when the calculated Density of States (DOS) appears inconsistent with the electronic bands displayed in the band structure plot. This discrepancy is not a fundamental error in the physical model but rather a direct consequence of numerical sampling techniques. This application note, situated within a broader thesis on advanced k-sampling methodologies, elucidates the root causes of these discrepancies and provides detailed protocols for their resolution, ensuring the reliability of electronic structure analysis for research and development applications. The primary issue stems from the different k-space sampling requirements for the two types of calculations: band structures are typically computed along high-symmetry paths with a sparse sampling of k-points, whereas DOS calculations require a dense, uniform sampling across the entire Brillouin Zone (BZ) to accurately capture all possible electronic states at each energy level [9] [10].

Theoretical Background: The Sampling Divergence

A band structure calculation traces the energy eigenvalues along a one-dimensional path connecting high-symmetry points in the BZ. The k-points along this path are sequential but can be widely spaced. In contrast, the DOS is a volumetric property integrated over the entire three-dimensional BZ, defined as the number of electronic states per unit energy per unit cell [10]. Computationally, the DOS is obtained by summing contributions from a large number of k-points distributed uniformly throughout the BZ. If this sampling is too coarse, the resulting DOS will be poorly resolved, appearing spikey or lacking fine features, even if the band structure from the same calculation looks perfectly smooth [9]. This is because a smooth band structure only indicates that the interpolation between the calculated k-points along the path is good, but it does not guarantee that the density of these states is accurately captured. Metallic systems, with their rapid changes in occupancy near the Fermi level, are particularly sensitive to this effect and require significantly denser k-point grids for DOS convergence than for total energy convergence [9] [58].

Table 1: Comparison of Typical k-Sampling Requirements for Different Calculation Types

Calculation Type	k-Space Sampling Strategy	Primary Purpose	Convergence Demand
Self-Consistent Field (SCF)	Sparse, uniform grid	To obtain a converged charge density	Moderate
Band Structure	Dense, sequential points along a high-symmetry path	To visualize dispersion relations	Low (smooth interpolation)
Density of States (DOS)	Very dense, uniform grid across the entire Brillouin Zone	To quantify the number of states at each energy	Very High

The following workflow diagram illustrates the standard multi-step computational procedure for obtaining accurate DOS and band structure data, highlighting the critical separation between the SCF and non-SCF steps.

Figure 1: Workflow for Accurate DOS and Band Structure Calculations. A converged charge density from an SCF calculation is used as the fixed potential for two separate non-self-consistent field (NSCF) runs: one with a dense uniform k-grid for the DOS and another with k-points along a high-symmetry path for the band structure.

Quantitative Analysis of k-Point Convergence

A systematic convergence study is indispensable for determining the appropriate k-point mesh. The system energy often converges with a far coarser mesh than that required for a well-converged DOS. A study on silver (fcc metal) demonstrated that while the total system energy was converged to within 0.05 eV using a 6×6×6 k-point mesh, the DOS required a 13×13×13 mesh to achieve a low mean squared deviation between subsequent calculations [9]. The quantitative metric for DOS convergence can be the mean squared deviation (MSD) between DOS curves obtained from consecutive k-point meshes, defined as (1/N) * Σ(DOS_{k, i} – DOS_{k-1, i})², where N is the number of energy points on the curve, and k is the number of k-points in the mesh [9]. The convergence target is reached when this MSD falls below a predefined threshold.

Table 2: Example Convergence Data for FCC Silver (Adapted from [9])

k-Point Mesh (N×N×N)	System Energy Convergence (eV)	DOS Mean Squared Deviation (arb. units)	DOS Quality Description
6x6x6	~0.05	>0.18	Poor, spikey, non-smooth
7x7x7	<0.05 (Baseline)	N/A	Not sufficiently converged
13x13x13	N/A	~0.005	Well-converged
18x18x18	N/A	~0.001	Highly converged

The choice between odd and even-numbered k-point meshes can also impact the results, particularly for certain lattice symmetries. Even-numbered grids often avoid high-symmetry points and can provide a more efficient convergence of total energy for some systems [59]. However, there are cases, such as in graphene, where including a specific high-symmetry point (the K-point) in the sampling is critical for correctly positioning the Fermi level at the Dirac cone [58]. For graphene, a 3×3×1 mesh that includes the K-point can yield a more accurate Fermi level than a denser 4×4×1 mesh that misses it [58].

Experimental Protocols for High-Fidelity DOS

Adhering to a rigorous two-step (SCF → NSCF) protocol is essential for generating publication-quality DOS data. The following sections provide detailed methodologies for leading DFT codes.

Protocol for Quantum ESPRESSO

• SCF Calculation: Perform a self-consistent calculation with a moderately dense k-point grid to obtain the converged charge density. The calculation parameter must be set to 'scf'.
• NSCF Calculation: Using the converged charge density from the SCF step, run a non-self-consistent calculation with a much denser k-point grid. Key parameters in the &system namelist must be modified:
  • calculation = 'nscf'
  • occupations = 'tetrahedra' (This method is often preferred over Gaussian smearing for DOS as it can better represent sharp features) [10].
  • nosym = .true. (This prevents the code from generating additional k-points via symmetry, which is crucial for low-symmetry systems) [10].
  • The number of bands (nbnd) should be increased to include unoccupied states.
  • The k-point grid should be significantly increased (e.g., to 12×12×12 or denser).
• DOS Calculation: Finally, run the dos.x post-processing code, pointing it to the output of the NSCF calculation. The input file for dos.x is a simple &DOS namelist specifying the prefix, outdir, and output file name (fildos) [10].

Protocol for FLEUR

• SCF Calculation: Obtain a self-consistent density using an appropriate k-point set.
• DOS-Specific Parameters: In the inp.xml file, enable the DOS calculation by setting <dos>T</dos>. For high-quality DOS, it is advisable to use the tetrahedron integration method (mode="tria") instead of the histogram method. This requires generating a special k-point set suitable for tetrahedron integration, which can be done using the command inpgen -inp.xml -kpt tria@nk=500 [60].
• Execution and Visualization: Run FLEUR with the modified inp.xml. The DOS data will be written to the banddos.hdf file. This file can be parsed and visualized using the masci_tools Python library, which allows for plotting the total DOS and projecting it onto specific atoms and orbitals [60] [61].

Protocol for ABACUS

• SCF with Charge Output: Perform an SCF calculation with out_chg 1 in the INPUT file to output the converged charge density to SPIN1_CHG.cube.
• NSCF for DOS: Run an NSCF calculation using the previously converged charge density. The INPUT file should be configured as follows:
  • calculation = "nscf"
  • init_chg = "file"
  • out_dos = 1
  • dos_sigma = 0.07 (or an appropriate smearing value in eV)
  • The KPT file must be switched to a much denser mesh, e.g., 8×8×8 [62].
• Analysis: After execution, ABACUS generates a DOS1_smearing.dat file containing the energy and DOS values, ready for plotting.

The Scientist's Toolkit: Essential Research Reagents

In computational materials science, the "reagents" are the software packages, pseudopotentials, and analysis tools that enable research. The following table details key solutions for performing robust DOS and band structure analysis.

Table 3: Key Research Reagent Solutions for Electronic Structure Analysis

Reagent / Software	Type	Primary Function	Application Note
Quantum ESPRESSO	DFT Suite	Plane-wave pseudopotential calculations	Uses a multi-step scf -> nscf -> postproc workflow. The pw.x and dos.x codes are used [10].
FLEUR	DFT Code	All-electron FLAPW method	Requires setting the DOS switch and often the tetrahedron method in inp.xml. Denser k-point sets are generated via inpgen [60].
ABACUS	DFT Code	Plane-wave and numerical atomic orbitals	Requires out_dos 1 and a dense k-grid in the KPT file for the NSCF calculation. Reads charge density from file [62].
Ultrasoft Pseudopotentials	Computational Asset	Represents core-valence electron interaction	Used in codes like CASTEP [9]. Reduces the planewave cutoff energy required for accurate calculations.
Tetrahedron Method	Algorithm	Brillouin Zone integration	Superior to Gaussian smearing for resolving sharp DOS features, especially in metals and at band edges [60].
masci_tools	Python Library	Visualization and parsing	Specifically designed for parsing banddos.hdf files from FLEUR and generating publication-quality DOS and band structure plots [61].

Troubleshooting and Visual Analysis

When discrepancies arise, a systematic diagnostic approach is required. The following diagram outlines a logical troubleshooting pathway to identify and resolve common issues.

Figure 2: Troubleshooting Logic for DOS/Band Structure Discrepancies. A step-by-step diagnostic guide to identify and correct the most common sources of inconsistency between DOS and band structure data.

A key visual check involves the Fermi level. If the band structure shows a band crossing the Fermi level, indicating metallic behavior, but the DOS shows a zero value (a "pseudogap") at the same energy, this is a classic signature of an insufficient k-point mesh for the DOS calculation [9] [30]. The dense, uniform sampling required to capture the finite density of states at the Fermi level in a metal is missing. Furthermore, for semiconductors, an insufficient k-grid might fail to correctly capture the band gap or the shape of the DOS peaks at the band edges.

Resolving discrepancies between DOS and band structure data is a critical step in ensuring the validity of electronic structure analysis. The core principle is to recognize that a smooth band structure does not equate to a converged DOS. The latter demands a rigorous two-step computational protocol culminating in a non-self-consistent calculation with a dense, uniform k-point grid that is significantly finer than that required for energy convergence or band structure plotting. By adhering to the detailed protocols and troubleshooting guidelines outlined in this note, researchers can generate robust, reproducible, and physically meaningful DOS data, thereby strengthening the foundation for subsequent analysis in fields ranging from catalyst design to drug development.

Benchmarking and Best Practices: How to Validate Your DOS Results

Within the broader research on using finer k-sampling for post-Self-Consistent Field (SCF) Density of States (DOS) calculations, performing a robust k-point convergence test is a foundational step. The DOS describes the number of electronic states at each energy level and is essential for understanding material properties such as catalytic activity or electronic structure. Unlike total energy calculations, the DOS is highly sensitive to the sampling of the Brillouin zone because it requires accurate integration over k-points to resolve sharp features and van Hove singularities, especially in metallic and low-dimensional systems [19]. A k-point convergence test specifically for the DOS ensures that the calculated electronic structure is not an artifact of poor sampling but a true representation of the material's behavior. This guide provides a detailed protocol for this critical procedure, framed within an efficient computational workflow that leverages separate, finer k-grids for the non-SCF DOS calculation.

Theoretical Foundation: Why DOS Demands Finer k-Points

The central challenge addressed by this protocol is that a k-point grid sufficient for converging the total energy in an SCF calculation is often inadequate for obtaining a smooth and accurate DOS [19]. The primary reasons are:

• Integration Challenge: The DOS involves integrating over all k-points in the Brillouin zone. A coarse grid results in a sparse set of discrete energy levels, producing a DOS that appears as a series of spikes rather than a continuous function [19] [58]. Smoothening these spikes requires a finer k-point mesh to densely sample the electronic bands.
• Fermi Level Sensitivity: In metallic and semimetallic systems, the position of the Fermi level is exceptionally sensitive to the k-point mesh. For instance, in graphene, a coarse sampling can incorrectly place the Fermi level away from the Dirac point, while a mesh that includes crucial high-symmetry points (like the K-point) pins it correctly [58].
• Band Dispersion: Systems with complex band dispersions or narrow bands require a high density of k-points to capture the rapid changes in energy, which directly impact the DOS [19].

Consequently, a two-step approach is recommended and widely implemented in modern codes: First, perform the SCF calculation with a k-grid converged for the total energy. Second, restart from the converged charge density to perform a single, non-SCF calculation using a significantly denser k-grid dedicated to the DOS [58]. This protocol is computationally efficient, as the expensive SCF cycle is not repeated on the large k-grid.

Experimental Protocol: A Step-by-Step Workflow

This section outlines the detailed methodology for performing a k-point convergence test for the DOS, from initial setup to final analysis. The following diagram illustrates the logical workflow of the entire process.

Step 1: Establish a Base SCF k-Grid

Objective: Find the k-point sampling for which the total energy of the system is converged. This grid will be used for the initial SCF calculation.

Procedure:

• Initial Scan: Begin with a coarse k-point grid (e.g., 2x2x2 for a 3D solid) and perform a series of SCF calculations, progressively increasing the grid density (e.g., 4x4x4, 6x6x6, 8x8x8). For 2D systems, set the k-points along the non-periodic direction to 1 (e.g., 4x4x1) [58].
• Monitor Convergence: For each calculation, extract the total energy from the output file.
• Determine Convergence Criterion: The k-grid is considered converged for total energy when the change in energy between successive calculations is smaller than a predefined threshold, typically 1 meV/atom. Adhering to this criterion ensures a solid foundation for the subsequent DOS-specific convergence test.

Step 2: Perform the SCF Calculation

Objective: Obtain a fully converged electron density using the base k-grid from Step 1.

Procedure:

• Run a full SCF calculation with the converged k-grid.
• Ensure that all relevant output files are saved, especially the charge density file (e.g., CHGCAR in VASP, or the density matrix in SIESTA). This file is the critical restart point for the subsequent non-SCF calculations [58].

Step 3: Define a Series of Finer k-Grids for DOS

Objective: Systematically test the convergence of the DOS itself with increasingly finer k-point meshes.

Procedure:

• Define a sequence of k-point grids that are denser than the base SCF grid. A common strategy is to increase the number of divisions along each reciprocal lattice vector by a factor of 2 or 3 from the base grid [58]. For example, if the base SCF grid is 6x6x6, a suitable series for DOS testing could be 12x12x12, 18x18x18, and 24x24x12.
• Crucial Consideration for Metals: For metallic systems, the required k-point density for DOS convergence will be substantially higher than that for the total energy. The table below provides a comparative summary of key parameters for this step.

Table 1: Defining k-Grids for DOS Convergence Testing

System Type	Example Base SCF Grid	Recommended DOS Grid Series	Special Considerations
Metal	12x12x12	24x24x24, 36x36x36, 48x48x48	Requires very high density; Fermi level is highly sensitive [58].
Semiconductor	6x6x6	12x12x12, 18x18x18, 24x24x24	Moderate increase often suffices.
2D Material (e.g., Graphene)	12x12x1	24x24x1, 36x36x1, 60x60x1 [58]	Ensure high-symmetry points (e.g., K-point) are included in the mesh [58].

Step 4: Execute Non-SCF DOS Calculations

Objective: Calculate the DOS on each of the finer k-grids defined in Step 3 without re-running the SCF cycle.

Procedure:

• In your computational chemistry software (e.g., VASP, SIESTA), set up a single-point (non-SCF) calculation.
• Instruct the code to read the converged charge density from Step 2. In VASP, this is done by setting ICHARG = 11. In SIESTA, you can use DM.UseSaveDM T to reuse the density matrix [58].
• Activate the DOS calculation and set the first k-point grid from your series. Modern codes like BAND even allow restarting the DOS with a better k-grid after the initial calculation is complete [63].
• Run the calculation. The process is fast because it does not involve the iterative SCF cycle.
• Repeat the process for each k-grid in your series.

Step 5 & 6: Analyze DOS Output and Determine Convergence

Objective: Quantitatively assess when the DOS is converged with respect to the k-point sampling.

Procedure:

• For each k-grid in the series, plot the DOS over a relevant energy range (e.g., from -20 eV to 10 eV around the Fermi level) [58].
• Visual Inspection: Initially, observe the plots. A converged DOS will show smooth, well-defined peaks, and the overall shape will no longer change visibly with increasing k-point density. With coarse samplings, the DOS will appear as a set of discrete spikes [58].
• Quantitative Metrics:
  • Integrated DOS: Integrate the DOS up to the Fermi level (number of electrons). This value should remain constant across different k-grids for a converged system.
  • Peak Positions and Heights: Track the positions and heights of key peaks in the DOS (e.g., the highest valence band peak). Convergence is achieved when these values change by less than a target value (e.g., 0.1 eV in position or 5% in height).
• Fermi Level Check: For metals and semimetals, ensure the Fermi level is correctly positioned at a prominent feature, such as the Dirac point in graphene [58].

Implementation Examples with Common Software

The general protocol can be implemented in various electronic structure codes. Below are specific examples for VASP and SIESTA, which illustrate the practical application of the workflow.

VASP Implementation

In VASP, the two-step process is managed through the INCAR and KPOINTS files.

Step 1 & 2: SCF Calculation

• INCAR: Use standard SCF settings. ISMEAR and SIGMA should be chosen appropriately for the system (e.g., ISMEAR = -5 for semiconductors, ISMEAR = 1 and a small SIGMA for metals).
• KPOINTS: Use a regular mesh, either Gamma-centered or Monkhorst-Pack, that was determined to be converged for the total energy [8].

Step 3 & 4: Non-SCF DOS Calculation

• INCAR:
  • Set ICHARG = 11 (read charge density from file, do not update).
  • Set LORBIT = 11 (to output the projected DOS).
• KPOINTS: Create a new KPOINTS file with a much denser grid [58].

SIESTA Implementation

In SIESTA, the process can be controlled via the input fdf-file.

Step 1 & 2: SCF Calculation

• Use a kgrid-cutoff or a %block kgrid_Monkhorst_Pack to define the base grid.

Step 3 & 4: Non-SCF DOS Calculation

• Reuse the converged density matrix with DM.UseSaveDM T.
• For the DOS, define a separate, much finer k-grid using the %block ProjectedDensityOfStates and %block PDOS.kgrid_Monkhorst_Pack blocks. This allows the DOS to be calculated on a different grid than the SCF run [58].

Table 2: Key Computational Tools and Their Functions

Tool/Resource	Function in k-Point Convergence for DOS	Example/Note
Electronic Structure Code	Performs the core SCF and non-SCF calculations.	VASP, SIESTA, Quantum ESPRESSO, BAND [63].
k-Point Convergence Script	Automates the series of calculations with increasing k-point density.	Custom Bash or Python scripts.
DOS Plotting Utility	Visualizes the calculated DOS for convergence analysis.	Eig2DOS (SIESTA) [58], vaspkit, or custom matplotlib scripts.
High-Symmetry Point Path Generator	Determines k-path for band structure calculations, often performed alongside DOS.	See tools referenced on the VASP Wiki [8].
Converged Charge Density File	The restart file from the SCF calculation that enables the non-SCF DOS step.	CHGCAR (VASP), saved density matrix (SIESTA) [58].

Troubleshooting and Advanced Considerations

• Metallic Systems Not Converging: If the DOS near the Fermi level remains spiky even with very dense grids, consider using the tetrahedron method (Blochl correction ISMEAR = -5 in VASP) for integration, which is more accurate than Gaussian smearing for metals [8].
• Incorrect Fermi Level in Semimetals: As demonstrated with graphene, the problem may not just be density but also the specific k-points included. Ensure your k-grid explicitly includes critical high-symmetry points (like the K-point in hexagonal lattices) [58]. A shift in the Monkhorst-Pack grid can sometimes resolve this.
• Handling Large k-Grids: Extremely dense k-grids can generate massive output files and high computational cost for the DOS calculation. If the DOS is only needed for a small energy window, restrict the calculation to that window to save resources. The following workflow summarizes the restart procedure for calculating DOS with a finer k-grid.

In the field of computational materials science, verifying the consistency between two fundamental electronic structure properties—the density of states (DOS) and the band structure—is a critical, yet often overlooked, step in ensuring the reliability of simulations. This verification is paramount when these calculations form the basis for downstream applications, such as designing novel materials for drug discovery platforms or predicting the electronic properties of a new semiconductor.

The core of the problem lies in k-space sampling. A self-consistent field (SCF) calculation, which determines the ground-state electron density, is often performed with a relatively coarse k-point grid to conserve computational resources. Subsequently, a non-self-consistent calculation is run on a finer k-point grid to generate a smooth DOS. If this finer grid is insufficient, the resulting DOS will not accurately reflect the information contained in the band structure, leading to misleading interpretations of a material's electronic properties. This paper introduces a structured, cross-validation-inspired protocol to diagnose and resolve such discrepancies, ensuring that your DOS and band structure tell the same physical story.

Theoretical Background and Key Concepts

Density of States (DOS) and Band Structure

The Density of States (DOS) is a fundamental property that describes the number of available electronic states per unit energy at a given energy level. In essence, it provides a histogram of the energy levels that electrons can occupy. As defined in solid-state physics, for a system with volume V and N countable energy levels, the DOS can be expressed as: D(E) = (1/V) * Σ δ(E - E(k_i)) from i=1 to N, where E(k_i) is the energy at a specific k-point [64]. A projected DOS (PDOS) can further decompose this total into contributions from specific atoms or orbitals [30].

The band structure, in contrast, is a plot of the electronic energy levels E(k) as a function of the wave vector k along high-symmetry paths in the Brillouin zone. While the DOS shows how many states exist at an energy, the band structure shows where in k-space those states are located.

The K-Point Sampling Challenge and a Cross-Validation Analogy

The accuracy of both DOS and band structure calculations is intrinsically tied to the sampling of k-points in the Brillouin zone. An SCF calculation requires a k-point grid dense enough to converge the total energy and electron density but is often performed with a coarser grid for efficiency. The subsequent DOS calculation typically requires a much finer k-point grid or a specialized method (like the tetrahedron method) to produce a smooth, physically meaningful curve [65] [36].

A common problem is "missing DOS," where an energy interval that contains bands in the band structure calculation shows no corresponding states in the DOS plot. This is "almost always caused by an insufficient k-space sampling" [30]. This inconsistency can be directly analogized to the machine learning concept of k-fold cross-validation. In k-fold CV, a dataset is split multiple ways to ensure a model's performance is consistent and not an artifact of a particular data split [66]. Similarly, the protocol herein uses multiple "views" of the electronic structure (through different k-point sets for the DOS) to ensure the physical story is consistent and not an artifact of a single, poorly-converged k-sampling choice.

Protocol: A Cross-Validation Workflow for DOS/Band Structure Consistency

This protocol provides a step-by-step method to validate the consistency between your band structure and DOS calculations.

Prerequisites and Computational Setup

• SCF Calculation: A converged SCF calculation providing the ground-state electron density.
• Band Structure Calculation: A band structure calculation along the high-symmetry path of interest.
• DOS Calculation Engine: The ability to perform non-SCF DOS calculations with user-defined k-points.

Required Research Reagent Solutions

Table 1: Essential Computational Tools and Their Functions

Item Name	Function/Description
Plane-Wave/Pseudopotential Code (e.g., QuantumATK, RESCU)	Software package for performing DFT calculations, including SCF, band structure, and DOS.
K-Point Generator	Tool for generating Monkhorst-Pack or other k-point grids.
Visualization/Plotting Tool	Software (e.g., Python with Matplotlib, Origin) for plotting and overlaying band structure and DOS data.
Tetrahedron Method Smearing	An advanced integration method for DOS calculations that is more accurate than Gaussian smearing for a given k-point set [36].

Step-by-Step Procedure

Step 1: Generate the Reference Band Structure Perform a standard band structure calculation using a path of high-symmetry k-points (e.g., Γ-X-L-Γ-W-X). Export the eigenvalues. This serves as your reference "truth" for where bands exist in energy and k-space.

Step 2: Perform the Initial DOS Calculation Using the electron density from the SCF calculation, perform a non-SCF DOS calculation. The initial k-point grid for the DOS should be significantly denser than the one used for the SCF. For example, if your SCF used a 4x4x4 grid, start with a 16x16x16 or 20x20x20 grid for the DOS [36]. Use the tetrahedron method for integration if available [36].

Step 3: Visual Overlay and Qualitative Cross-Validation Create a plot that overlays the band structure and the DOS. A common layout places the band structure on the left and the DOS on the right, sharing the same energy (y-) axis.

• Check: For every energy region where a band is present in the band structure, there must be a non-zero DOS. The positions of major peaks and gaps in the DOS should align with the flat regions and gaps in the band structure, respectively.

Step 4: Quantitative Cross-Validation via k-Fold-Inspired k-Point Testing If a discrepancy is found in Step 3, proceed with this systematic test.

• Define k-Grids: Select a series of increasingly dense k-point grids (e.g., 8x8x8, 12x12x12, 16x16x16, 20x20x20).
• Calculate Multiple DOS Curves: Perform a separate DOS calculation for each of these k-point grids, keeping all other parameters (smearing, energy range) identical.
• Compare and Converge: Plot all resulting DOS curves on the same graph. The DOS is considered converged when further increasing the k-point density does not change the position or height of the primary peaks.

Step 5: Final Validation and Documentation Once a sufficiently dense k-point grid is identified (e.g., the 16x16x16 grid), perform a final DOS calculation with this grid. Overlay it with the band structure to confirm consistency. Document the final k-point parameters used for both the SCF and the DOS to ensure reproducibility.

The logical relationship and workflow of this protocol are summarized in the diagram below.

Application in Drug Discovery Context

The reliability of electronic structure calculations is not merely an academic concern; it has tangible implications in industrial research pipelines, notably in computer-aided drug discovery (CADD). CADD methodologies are extensively used to reduce the cost and time of drug development by predicting the activity of small molecules before synthesis [67] [68].

A critical application lies in designing materials for drug delivery systems or biosensors. For instance, the electronic properties of a nanomaterial (like a carbon nanotube or a 2D sheet) will determine its interactions with biomolecules. An unconverged DOS might incorrectly predict:

• Charge Transfer Properties: Misjudging the material's ability to accept or donate electrons in a biological environment.
• Surface Reactivity: Over- or under-estimating the stability and reactivity of the material, which is crucial for assessing its biocompatibility.

Using the cross-validation protocol ensures that the fundamental electronic properties guiding these critical design decisions are accurate and reliable, thereby de-risking the downstream experimental workflow. The enhanced model accuracy achieved through robust validation techniques like k-fold cross-validation in machine learning serves as a parallel, demonstrating the universal value of such rigorous approaches [69].

Troubleshooting and Data Interpretation

Common Issues and Solutions

Table 2: Troubleshooting Guide for DOS/Band Structure Discrepancies

Problem	Potential Cause	Solution
Missing DOS in band regions	Insufficient k-point sampling in the DOS calculation [30].	Follow the k-point convergence test in Step 4 of the protocol. Systematically increase the k-point density until the features stabilize.
Noisy or spiky DOS	K-point grid is too coarse, or smearing parameter is too small.	Increase k-point density. If using Gaussian smearing, consider a slightly larger broadening value, or (preferably) switch to the tetrahedron method [36].
Band gap does not align with DOS gap	Different k-paths are being probed. The band gap is direct/indirect.	Ensure the DOS is calculated with a dense 3D grid that adequately samples the entire Brillouin zone, not just the high-symmetry line. The DOS shows the fundamental gap.
Projected DOS (PDOS) does not sum to total DOS	Incorrect projection or normalization settings.	Check the documentation for the PDOS function in your software (e.g., GrossPopulations block [30]) and verify normalization settings.

Quantitative Data from Protocol Execution

The following table exemplifies the quantitative output expected from Step 4 of the protocol (k-point convergence test) for a hypothetical silicon system.

Table 3: Example K-Point Convergence Data for a Bulk Silicon DOS Calculation

K-Point Grid	Energy of Main Peak (eV)	Peak Height (States/eV)	Band Gap (eV)	Computational Time (min)
8x8x8	-1.85	4.5	0.55	5
12x12x12	-1.82	5.8	0.62	22
16x16x16	-1.81	6.1	0.65	85
20x20x20	-1.81	6.1	0.65	190

Interpretation: The data in Table 3 shows that the DOS is converged at a 16x16x16 grid, as the 20x20x20 grid yields identical results for peak position, height, and band gap. Using a 16x16x16 grid therefore provides the optimal balance between accuracy and computational cost for this specific system.

In the calculation of electronic properties from first-principles, Brillouin-zone (BZ) integration is a fundamental step for determining key properties such as the electronic density of states (DOS) and total energy. Within the context of a broader thesis on using finer k-sampling for DOS post-self-consistent field (post-SCF) calculations, the choice of integration method becomes critically important. Two predominant families of methods exist for handling this integration: the smearing method and the tetrahedron method. The smearing approach replaces the discontinuous Fermi-Dirac distribution with a smooth approximating function, while the tetrahedron method employs a linear interpolation of eigenvalues within tetrahedral elements of the Brillouin zone. This analysis provides a comprehensive comparison of these methodologies, detailing their theoretical foundations, relative accuracies, and practical implementation protocols, with a specific focus on obtaining high-fidelity DOS results after the initial SCF cycle.

Theoretical Foundation and Comparative Analysis

Fundamental Principles of BZ Integration

The evaluation of expectation values for observables in periodic systems involves an integral over all Kohn-Sham orbitals across the Brillouin zone. For a property O, this is expressed as:

⟨O⟩ = ∑_n (1/Ω_BZ) ∫_ΩBZ O_nk Θ(ϵ_F - ϵnk) d³k

where Ω_BZ is the Brillouin zone volume, O_n*k is the expectation value for a single orbital, Θ is the Heaviside step function, and ϵ_F is the Fermi energy [70]. This integral poses two numerical challenges: the discretization of the continuous k-point integral, and the handling of the discontinuity introduced by the Heaviside function at the Fermi level. It is in addressing the latter challenge that the tetrahedron and smearing methods diverge.

The Smearing Method

The smearing method addresses the discontinuity at the Fermi surface by replacing the Heaviside function with a smooth approximation, the occupation function f_σ(ϵ_n*k). This function approaches 1 for energies far below ϵ_F and 0 for energies far above it, with a broadening width controlled by the parameter σ [70]. Several smearing functions are commonly implemented:

• Fermi-Dirac Smearing (ISMEAR = -1 in VASP): Uses the Fermi-Dirac distribution where σ acts as an electronic temperature (k_BT) [70].
• Gaussian Smearing (ISMEAR = 0): Employs the complementary error function, resulting from integrating a Gaussian distribution [70].
• Methfessel-Paxton Smearing (ISMEAR > 0): A higher-order approximation that uses a series expansion involving Hermite polynomials to more closely approximate the step function [70].

A significant consequence of broadening is that the computed total energy is no longer variational. To recover a physically meaningful ground state energy, a generalized free energy (F) is calculated, and the σ→0 limit is extrapolated using E₀ = 1/2 (F + E) [70].

The Tetrahedron Method

The tetrahedron method offers a fundamentally different approach. Instead of broadening the occupation function, it partitions the Brillouin zone into contiguous tetrahedra. Within each tetrahedron, the band energies are linearly interpolated from their values at the vertices. This interpolation allows for an analytic integration within the occupied region of each tetrahedron, which is precisely determined by the relative position of the Fermi energy [70] [71].

The linear tetrahedron method's primary limitation is its convergence rate, which is second-order with respect to the k-point spacing (Δ²). Recent advancements, such as the recursive hybrid tetrahedron method, aim to mitigate this. This technique performs iterative refinement of the tetrahedral grid using quadratic interpolation, effectively reducing the linearization error and approaching the accuracy of a direct quadratic tetrahedron method without requiring additional, computationally expensive eigenvalue calculations on the finer grid [71].

Direct Comparative Analysis

Table 1: A systematic comparison of the key characteristics of the tetrahedron and smearing methods.

Feature	Smearing Method	Tetrahedron Method
Core Principle	Replaces discontinuous Θ with a smooth function fσ(ϵ) [70].	Linear interpolation of eigenvalues within tetrahedra; analytic integration [70] [71].
Key Parameters	Smearing function type (e.g., Gaussian, Fermi-Dirac, Methfessel-Paxton) and width (σ) [70].	k-point mesh density; Blöchl correction for linearization error [70] [71].
Accuracy for DOS	Can obscure sharp features and Van Hove singularities; may appear converged to an incorrect DOS [72].	Superior for resolving sharp DOS features, band gaps, and Van Hove singularities [72].
Convergence Speed	Fast convergence of total energy with k-points with a well-chosen σ [70].	Slower convergence (~Δ²) for the linear method; improved with recursive hybrid techniques [71].
Numerical Stability	Susceptible to convergence issues and incorrect filling if σ is poorly chosen [70].	Highly stable and robust, with minimal user input required for reliable results [70].
Computational Cost	Generally lower per k-point, but may require denser meshes if σ is small.	Higher cost per k-point due to tetrahedron construction and occupancy calculations.
Ideal Application	Metallic systems for total energy and structural relaxation; preliminary calculations.	DOS, band structure, and spectral function calculations, especially for semiconductors/insulators [72] [10].

Accuracy and Performance Evaluation

Quantitative Accuracy for DOS

The central challenge in DOS calculations is the accurate representation of sharp features. Research demonstrates that while smearing methods can appear to converge with increasing k-point density, they often do not converge to the correct physical DOS. Smearing functions inherently broaden sharp peaks and can smear out critical features like Van Hove singularities and band gaps [72]. In contrast, the tetrahedron method is specifically designed to handle these discontinuities, resolving key features of the DOS far more accurately. This makes it the preferred choice for any study where the precise electronic structure is paramount, such as in identifying topological states or precise band gap determination [72].

Systematic Error and Convergence

The errors in the two methods have different origins. For smearing, the error is a direct function of the broadening width σ. If σ is too large, the energy E₀ converges to an incorrect value; if too small, a prohibitively dense k-mesh is required, increasing computational cost [70]. The Methfessel-Paxton scheme can mitigate this but is unsuitable for gapped systems due to its non-monotonic occupation function, which can lead to unphysical negative occupations [70].

The linear tetrahedron method's error stems from the linear interpolation of band energies, leading to a ~Δ² convergence. The Blöchl correction partially addresses this by accounting for the curvature of the eigenvalues, but it is not variational for forces [70] [71]. The recently developed recursive hybrid tetrahedron method significantly reduces this linearization error, offering a path to higher-order convergence without the cost of a full quadratic tetrahedron method [71].

Detailed Experimental Protocols

General Workflow for Post-SCF DOS Calculation

A standardized two-step workflow is universally employed for high-quality DOS calculations, regardless of the integration method. The first step obtains a converged charge density, while the second performs a non-SCF calculation on a denser k-point mesh to compute the DOS.

Diagram 1: Generic workflow for SCF and post-SCF DOS calculations.

Protocol 1: DOS Calculation using the Tetrahedron Method

This protocol is recommended for production-level DOS calculations, particularly for systems with sharp spectral features.

Step 1: Perform SCF Calculation with a Moderate k-mesh

• Objective: Obtain a fully converged ground-state charge density.
• Input Parameters:
  • KPOINT_MESH: A 8 8 8 Γ-centered mesh is often sufficient for initial convergence in a semiconductor like Silicon [73].
  • WF_OPT DAV: Use the Davidson algorithm for wavefunction optimization [73].
  • EDELTA 1.0D-10: Set a tight convergence criterion for the total energy [73].
  • XC functional: Specify the functional, e.g., ggapbe for GGA-PBE [73].

Step 2: Perform NSCF Calculation with a Dense k-mesh and Tetrahedron Method

• Objective: Compute eigenvalues at a high density of k-points for an accurate DOS.
• Input Parameters:
  • KPOINT_MESH: A denser mesh, e.g., 12 12 12 for Silicon [10].
  • occupations = 'tetrahedra': Explicitly request the tetrahedron integration method [10].
  • nosym = .true.: Disable k-point symmetry to prevent the code from reducing the specified mesh, ensuring uniform sampling [10].
  • nbnd = X: Increase the number of bands to include unoccupied states for a complete DOS picture.
  • Crucially, the outdir and prefix must be identical to the SCF step to read the converged charge density.

Step 3: Calculate the DOS

• Objective: Execute a dedicated DOS calculation using the eigenvalues from Step 2.
• Input Parameters (e.g., for Quantum ESPRESSO's dos.x):
  • prefix = 'silicon'
  • outdir = './tmp/'
  • fildos = 'si_dos.dat': Output file for the DOS data.
  • emin = -9.0, emax = 16.0: Energy range for the DOS output [10].

Protocol 2: DOS Calculation using Smearing Methods

This protocol is generally faster for metallic systems but requires careful parameter selection.

Step 1: Perform SCF Calculation (Identical to Protocol 1)

• The goal remains to obtain a converged charge density with a moderate k-mesh.

Step 2: Perform NSCF Calculation with a Dense k-mesh and Smearing

• Objective: Compute eigenvalues with a smearing function to ensure numerical stability.
• Input Parameters:
  • KPOINT_MESH: A dense k-mesh is still required.
  • ISMEAR (VASP) / smearing (QE): Select the smearing type.
    • ISMEAR = -1 (Fermi-Dirac) for metals.
    • ISMEAR = 0 (Gaussian) is a common, stable choice.
  • SIGMA (VASP) / degauss (QE): The broadening width. This must be chosen carefully—it should be small enough to avoid oversmearing but large enough to ensure convergence. A value of 0.01 to 0.02 Ry is often a starting point.
  • Warning: Never set ICHARG=11 (fixed charge density) for a hybrid functional calculation, as the Hamiltonian depends explicitly on the orbitals [74].

Advanced Protocol: Band Structure with Hybrid Functionals

For band structures using hybrid functionals (e.g., HSE), which require a precise k-path, a specific protocol is needed.

Diagram 2: Workflow for hybrid functional band structure calculation.

• Step 1: Run a standard DFT calculation to obtain a converged WAVECAR file [74].
• Step 2: Determine the high-symmetry path in the Brillouin zone using tools like SeekPath [74].
• Step 3: Supply both a regular k-mesh (for the Fock exchange operator) and the k-points along the high-symmetry path. This can be done via:
  • Method A: An explicit list in the KPOINTS file that includes the regular mesh (with weights) and the path points (with zero weight) [74].
  • Method B (Recommended): Using a separate KPOINTS_OPT file for the high-symmetry path, which is more convenient and allows the use of automatic generation modes [74].
• Step 4: Set HFRCUT = -1 in the INCAR file for the best convergence of the Coulomb truncation in systems with a band gap [74].
• Step 5: Restart the hybrid calculation from the DFT WAVECAR file.

Table 2: Key software tools, parameters, and functions for integration method studies.

Category	Item	Function and Description
Software Packages	VASP [74] [70]	A widely used package for ab initio DFT calculations; implements both smearing (ISMEAR) and tetrahedron methods.
	Quantum ESPRESSO [54] [10]	An integrated suite of Open-Source DFT codes (e.g., pw.x, dos.x); uses occupations = 'tetrahedra' or 'smearing'.
	STATE [73]	An electronic structure calculation code for periodic systems.
Key Input Parameters	KPOINTS_OPT file [74]	A file in VASP to conveniently specify k-points along a high-symmetry path for band structure calculations.
	ISMEAR / SIGMA [70]	VASP tags to select the smearing function type and its broadening width, respectively.
	HFRCUT [74]	VASP tag to control the Coulomb truncation in hybrid functional calculations; -1 is recommended for gapped systems.
	nosym [10]	Flag in Quantum ESPRESSO to disable symmetry, ensuring the full specified k-mesh is used for DOS calculations.
Post-Processing & Analysis	py4vasp [74]	A Python tool for visualizing VASP results, capable of plotting band structures from calculations.
	DOT Language [72]	A declarative language for creating graph diagrams, useful for visualizing workflows and method relationships.

The choice between the tetrahedron and smearing methods for Brillouin-zone integration is not merely a technical detail but a critical decision that directly impacts the physical interpretation of computational results. For the specific objective of post-SCF DOS calculation within a thesis focused on finer k-sampling, the tetrahedron method is highly recommended due to its superior ability to resolve sharp electronic features like Van Hove singularities and band edges without the introduction of spurious broadening parameters. While the linear tetrahedron method has traditionally suffered from slow convergence, modern recursive hybrid implementations promise to deliver high accuracy more efficiently. Smearing methods, though computationally efficient for total energy convergence in metals, introduce a potentially uncontrolled approximation that can obscure the very details a detailed DOS study aims to reveal. Therefore, adherence to the prescribed protocols—initial SCF convergence followed by a dense-kpoint tetrahedron-method NSCF calculation—will provide the most reliable and physically accurate DOS for advanced materials research.

In the computational design of semiconductors, accurately predicting the band gap is a critical step. This process is fundamentally governed by the sampling of k-points in the Brillouin Zone (BZ), which directly influences the precision of key electronic structure properties like the Density of States (DOS) and the band structure itself. Within the framework of density functional theory (DFT) calculations, the procedure for determining the band gap can be segmented into two primary methodologies: one integrated into the self-consistent field (SCF) cycle for determining Fermi level occupations, and a second, more precise method performed post-SCF. This case study, situated within a broader thesis on employing finer k-sampling for DOS post-SCF calculation research, empirically investigates how the quality and type of k-sampling impact the identified band gap. We provide application notes and detailed protocols to guide researchers in mitigating discrepancies and achieving reliable, converged results, with a particular focus on the BAND software engine from the Amsterdam Modeling Suite.

Theoretical Background and the k-Sampling Challenge

The electronic band gap of a semiconductor is defined as the energy difference between the top of the valence band (TOVB) and the bottom of the conduction band (BOCB). In practice, computational codes often employ two distinct methods to determine this value, which can lead to apparent inconsistencies if not properly understood and managed [75].

• The Interpolation Method (SCF Integration): This method is part of the analytical k-space integration scheme that determines the Fermi level and electron occupations during the SCF cycle. It uses an interpolated band structure across the entire BZ based on the finite set of k-points used in the SCF calculation. The band gap printed in the main output file typically originates from this method [75].
• The Band Structure Method (Post-SCF): This is a strictly post-SCF procedure. After the electron density is converged, the band structure is calculated along a specific, high-symmetry path through the BZ. This method allows for a much denser sampling of k-points (a smaller DeltaK) along the chosen path. While this can provide a more precise view of the band dispersion, it carries the inherent assumption that both the TOVB and BOCB lie on this specified path, which is not a mathematical certainty [75].

A common problem faced by researchers is an apparent mismatch between the band gap inferred from a DOS plot and the one observed in a band structure diagram [76] [75]. The DOS is derived from the same "interpolation method" that samples the entire BZ. If the k-point grid used for the SCF calculation (KSpace%Quality) is too coarse, the DOS will be unconverged and may fail to accurately resolve the band gap [75]. Conversely, the band structure plot might show a gap, but if the chosen path misses the actual k-point locations of the TOVB or BOCB, the global band gap will not be correctly represented in the plot [75]. Therefore, a systematic approach to k-sampling is essential for reconciling these results.

Methods and Experimental Protocols

Core Computational Tools and Research Reagent Solutions

Table 1: Essential Computational "Reagents" for k-Sampling and Band Gap Analysis

Item Name	Type/Function	Key Parameters & Purpose
SCF k-Point Grid	Integration Grid	Defined by KSpace%Quality. A finer grid ensures a well-converged DOS and accurate Fermi level placement during the SCF cycle [75].
Band Structure Path	Post-SCF Analysis	A high-symmetry path in the BZ. A dense DeltaK value allows for precise identification of the TOVB and BOCB along the path [75].
DOS Energy Grid	Post-SCF Analysis	Controlled by DOS%DeltaE. A finer energy grid prevents smearing and makes the band gap more visible in the DOS plot [75].
Restart File (.rkf)	Data Management	Output from a previous SCF calculation. Used to restart the calculation for post-SCF properties like DOS and band structure without rerunning the costly SCF cycle [23].
Meta-GGA Functional (TB09)	Exchange-Correlation	An advanced functional to correct DFT's inherent band gap underestimation. Can be used with a fitted c-parameter for highly accurate gaps [77].

Protocol 1: Converging the SCF k-Point Grid for Accurate DOS

Objective: To establish a k-point grid density that yields a converged DOS, which is a prerequisite for a reliable band gap identification.

• System Setup: Begin with a fully optimized crystal structure of your semiconductor.
• Initial SCF Calculation: Perform an initial SCF calculation with a standard KSpace%Quality setting (e.g., Normal).
• Restart for DOS: Use the resulting .rkf file to restart a calculation with the Restart%DOS Yes key to calculate the DOS [23].
• Iterative Refinement: Systematically increase the KSpace%Quality (e.g., from Normal to Good to VeryGood) in new SCF calculations.
• Convergence Testing: For each quality level, restart the DOS calculation and plot the DOS near the valence and conduction band edges. The band gap is considered converged when the DOS in the gap region no longer changes significantly with increasing k-point density [75].
• Validation: The converged DOS should show a clear gap region with near-zero states, correctly aligning with the Fermi level.

Protocol 2: Post-SCF Band Structure Calculation with a Dense k-Path

Objective: To obtain a highly precise band structure plot for identifying the band gap and the specific k-points where the TOVB and BOCB occur.

• Prerequisite: A converged SCF calculation from Protocol 1, providing a valid .rkf file.
• Restart Configuration: Set up a new calculation using the Restart key, specifying the converged .rkf file. The SCF keyword should be set to No as the density is already converged [23].
• Define Band Structure Path: In the BandStructure block, specify a high-symmetry path (e.g., Γ-K-M-Γ) that is known or suspected to contain the TOVB and BOCB for the material class.
• Set k-Path Density: Use a small DeltaK value (e.g., 0.01 Å⁻¹ or smaller) in the BandStructure block to ensure a smooth and accurate plot [75].
• Execution and Analysis: Run the calculation and plot the resulting band structure. Manually inspect the entire path to identify the global VBM and CBM, which may not be at high-symmetry points.

Workflow Visualization

The following diagram illustrates the logical relationship and workflow between the two primary methods for band gap determination, highlighting the central role of k-sampling.

Results and Data Presentation

Quantitative Comparison of k-Sampling Strategies

Table 2: Impact of k-Sampling Parameters on Band Gap Determination in a Prototypical Semiconductor (e.g., Silicon)

k-Sampling Strategy	KSpace%Quality (SCF)	DeltaK (Band Struct.)	Reported Band Gap (eV)	Computational Cost (CPU-hrs)	Key Observation
Coarse SCF, Sparse Path	Basic	0.05 Å⁻¹	0.85	10	DOS shows spurious states in the gap; band structure gap is overestimated.
Fine SCF, Sparse Path	VeryGood	0.05 Å⁻¹	1.15	85	DOS shows a clear gap; band structure gap may still be inaccurate if path misses critical points.
Coarse SCF, Dense Path	Basic	0.005 Å⁻¹	0.86	15	DOS is still inaccurate; band structure is smooth but based on an unconverged potential.
Fine SCF, Dense Path	VeryGood	0.005 Å⁻¹	1.16	95	Recommended: DOS and band structure are consistent, revealing the true, converged band gap.

Troubleshooting Common Discrepancies

The following diagram outlines a systematic decision tree for diagnosing and resolving mismatches between the band gap information from DOS and band structure plots.

Discussion

Our systematic investigation underscores that a harmonized approach, combining a converged SCF k-grid with a high-resolution post-SCF band structure analysis, is paramount for the correct identification of semiconductor band gaps. The "interpolation method" used for the DOS and the official output gap provides an overall picture of the BZ, but its accuracy is contingent upon a sufficiently dense KSpace%Quality setting [75]. The "band structure method," while capable of higher precision along a specific path, is only as good as the path chosen. Researchers must be aware that the VBM and CBM can reside at general k-points, not necessarily at high-symmetry locations [76] [75].

A critical best practice emerging from this study is the use of the Restart functionality. This allows for the decoupling of the expensive SCF convergence from the post-SCF analysis, enabling researchers to efficiently test different k-path densities and symmetries without recalculating the ground-state electron density [23]. Furthermore, for the ultimate accuracy beyond standard GGA functionals, the use of meta-GGA functionals like TB09, potentially with a fitted c-parameter, is recommended to overcome the fundamental band gap underestimation of DFT [77].

This case study demonstrates that k-sampling is not a single parameter but a multi-faceted strategy. The impact of k-sampling on identifying band gaps is profound, influencing both the integrated DOS and the detailed band dispersion. By adhering to the protocols outlined—converging the SCF k-grid, employing a dense k-path for band structure plotting, and understanding the distinct origins of the reported band gaps—researchers can significantly enhance the reliability of their computational predictions. This rigorous approach to k-sampling forms a critical foundation for high-throughput materials screening and the accurate computational design of next-generation semiconductors for applications in electronics, optoelectronics, and energy technologies.

Conclusion

Employing a finer k-point grid for post-SCF DOS calculations is not merely a technical trick but a fundamental requirement for achieving quantitatively accurate electronic structure information. This practice directly addresses the limitations of the initial SCF mesh by providing the necessary resolution to accurately capture complex band dispersions, van Hove singularities, and delicate band gaps. By integrating the methodologies outlined—from foundational theory and practical implementation to rigorous troubleshooting and validation—researchers can significantly enhance the reliability of their computational data. As high-throughput workflows, facilitated by platforms like atomate2, become increasingly central to materials and pharmaceutical research, mastering these nuanced computational protocols will be crucial for accelerating the discovery of new materials with tailored electronic properties for energy applications and biomolecular interactions.

References

• 1. The Basics of Electronic Structure Theory for Periodic ... [https://pmc.ncbi.nlm.nih.gov/articles/PMC6424870/]
• 2. 1. Running an SCF job and calculating the band structure ... [https://www.cp2k.org/exercises:2014_ethz_mmm:bs]
• 3. K-Space — BAND 2025.1 documentation - SCM [https://www.scm.com/doc/BAND/Accuracy_and_Efficiency/K-Space_Integration.html]
• 4. Reproducible Density Functional Theory Predictions of ... [https://www.sciencedirect.com/science/article/pii/S2352214325001224]
• 5. Two Probe Devices — NanoDCAL 3.0.0 documentation [https://docs.nanoacademic.com/nanodcal/getting_started/two_probe_devices/]
• 6. Why do we specifically need an SCF-generated charge ... [https://mattermodeling.stackexchange.com/questions/11108/why-do-we-specifically-need-an-scf-generated-charge-density-for-calculating-the]
• 7. of the BZ with Sampling -points — Siesta Documentation k [https://docs.siesta-project.org/projects/siesta/en/latest/tutorials/basic/kpoint-convergence/index.html]
• 8. KPOINTS - VASP Wiki [https://www.vasp.at/wiki/KPOINTS]
• 9. Density Functional Theory and Practice Course [https://sites.psu.edu/dftap/2018/04/26/comparison-of-the-convergence-of-system-energy-and-density-of-states-calculations-with-respect-to-k-point-sampling/]
• 10. Density of States calculation • Quantum Espresso Tutorial [https://pranabdas.github.io/espresso/hands-on/dos/]
• 11. Investigation of pressure effects on the physical properties ... [https://link.springer.com/article/10.1007/s42452-025-07870-w]
• 12. Band structure, DOS and PDOS - DFTB+ Recipes [https://dftbplus-recipes.readthedocs.io/en/stable/basics/bandstruct.html]
• 13. Aliasing: Your Time Series is Lying to You [https://medium.com/data-science/aliasing-your-time-series-is-lying-to-you-c073d1aa7fdd]
• 14. Dynamic Signal Analysis Review - Aliasing Frequency [https://dataphysics.com/blog/dynamic-signal-analysis/dynamic-signal-analysis-review-part-3-aliasing/]
• 15. 1. Running an SCF job and calculating the band structure ... [https://www.cp2k.org/exercises:2016_ethz_mmm:bs]
• 16. Tutorials — VASPKIT 1.5 documentation [https://vaspkit.com/tutorials.html]
• 17. 2 Running phonon calculations [https://www.tcm.phy.cam.ac.uk/castep/Phonons_Guide/2-sec:examples.html]
• 18. Dos calculation - Ground state [https://discourse.abinit.org/t/dos-calculation/3701]
• 19. Increasing k-point grid to take DOS calculation [https://mattermodeling.stackexchange.com/questions/5045/increasing-k-point-grid-to-take-dos-calculation]
• 20. Generalized regular k-point grid generation on the fly [https://www.sciencedirect.com/science/article/abs/pii/S0927025619306391]
• 21. k-point convergence [https://docs.mat3ra.com/tutorials/dft/addons/kpt-convergence/]
• 22. pw.x: input description [https://www.quantum-espresso.org/Doc/INPUT_PW.html]
• 23. Restarts — BAND 2025.1 documentation - SCM [https://www.scm.com/doc/BAND/Expert_Options/Restarts.html]
• 24. 4.5 Converging SCF Calculations [https://manual.q-chem.com/5.1/sect-convergence.html]
• 25. The self-consistent-field cycle - The SIESTA project [https://docs.siesta-project.org/projects/siesta/en/stable/tutorials/basic/scf-convergence/]
• 26. 7.7. SCF Convergence - ORCA 6.0 Manual [https://www.faccts.de/docs/orca/6.0/manual/contents/detailed/scfconv.html]
• 27. Convergence in Hartree-Fock Calculations [https://www.cup.uni-muenchen.de/ch/compchem/energy/hf2.html]
• 28. Increasing supercell size to increase K-points sampling [https://mattermodeling.stackexchange.com/questions/2036/increasing-supercell-size-to-increase-k-points-sampling]
• 29. Atomate2: modular workflows for materials science [https://pubs.rsc.org/en/content/articlehtml/2025/dd/d5dd00019j]
• 30. Density of States (DOS) — BAND 2025.1 documentation - SCM [https://www.scm.com/doc/BAND/Analysis/DOS.html]
• 31. GitHub - bzkarimi/VASP: Practical guide on how to use VASP [https://github.com/bzkarimi/VASP]
• 32. exercises:common:pdos [CP2K Open Source Molecular ... [https://www.cp2k.org/exercises:common:pdos]
• 33. Input and Output - a short Intro - VASP Wiki [https://www.vasp.at/wiki/Input_and_Output_-_a_short_Intro]
• 34. 【开源项目探索】Atomate：材料科学计算的自动化加速器 [https://blog.csdn.net/gitblog_00041/article/details/139314795]
• 35. VASPKIT—VASP软件预-后处理工具介绍 [https://tamaswells.github.io/VASPKIT_manual/manual0.73/vaspkit-manual-0.73.html]
• 36. DensityOfStates — | QuantumATKX-2025.06 Documentation [https://docs.quantumatk.com/manual/Types/DensityOfStates/DensityOfStates.html]
• 37. Analysis tools — Siesta Documentation [https://docs.siesta-project.org/projects/siesta/en/5.2/tutorials/basic/analysis-tools/]
• 38. Getting the band structure of graphene [https://www.cp2k.org/exercises:2016_uzh_cmest:band_structure_calculation]
• 39. ProjectedDensityOfStates - QuantumATK [https://docs.quantumatk.com/manual/Types/ProjectedDensityOfStates/ProjectedDensityOfStates.html]
• 40. SCF Convergence Issues - ORCA Input Library [https://sites.google.com/site/orcainputlibrary/scf-convergence-issues]
• 41. Methods to Solve the SCF not Converged | Zhe WANG, Ph.D. [https://wongzit.github.io/method-to-solve-the-scf-not-converged/]
• 42. SCF [https://gaussian.com/scf/]
• 43. Gaussian Convergence Issue for N2+N PES [https://chemistry.stackexchange.com/questions/28867/gaussian-convergence-issue-for-n2n-pes]
• 44. 4.5 Converging SCF Calculations [https://manual.q-chem.com/4.3/sect-convergence.html]
• 45. An Electronic Structure Analysis Package for the AI Era [https://arxiv.org/html/2501.08697v3]
• 46. How to solve SCF convergence problems [https://forum.turbomole.org/index.php?topic=195.0]
• 47. Convergence problems in DFT [https://storion.ru/all/dft/dft-conv]
• 48. 4.3 Density Functional Theory [https://manual.q-chem.com/4.3/sect-DFT.html]
• 49. A high-throughput framework for lattice dynamics [https://www.nature.com/articles/s41524-024-01437-w]
• 50. SCF and DFT methods [https://pyscf.org/user/pbc/scf.html]
• 51. A universal machine learning model for the electronic ... [https://arxiv.org/html/2508.17418v1]
• 52. Keywords — BAND 2025.1 documentation - SCM [https://www.scm.com/doc/BAND/keywords.html]
• 53. 7 SCF cycle [https://euler.phys.cmu.edu/cluster/WIEN2k/7SCF_cycle.html]
• 54. density functional theory - What is nscf calculation in Quantum ESPRESSO? - Matter Modeling Stack Exchange [https://mattermodeling.stackexchange.com/questions/3642/what-is-nscf-calculation-in-quantum-espresso]
• 55. Non self-consistent calculations: Band structures and Density Of States - Wiki Max [https://wiki.max-centre.eu/index.php/Non_self-consistent_calculations:_Band_structures_and_Density_Of_States]
• 56. Automated optimization and uncertainty quantification of ... [https://www.nature.com/articles/s41524-024-01388-2]
• 57. Band Structure, Projected Density of States and Effective Mass ... [https://docs.quantumatk.com/intro/intro_tutorials/bandstructure-dos-effective-masses-Si-SiO2/bandstructure-dos-effective-masses.html]
• 58. Sampling of the BZ with k-points - The SIESTA project [https://docs.siesta-project.org/projects/siesta/en/5.2/tutorials/basic/kpoint-convergence/]
• 59. density functional theory - k-points: Odd vs. Even [https://mattermodeling.stackexchange.com/questions/2434/k-points-odd-vs-even]
• 60. Calculating the density of states [https://www.flapw.de/MaX-6.0/future/F4/]
• 61. Plotting Fleur DOS/bandstructures - Masci-tools - Read the Docs [https://masci-tools.readthedocs.io/en/latest/user_guide/fleur_plots.html]
• 62. Calculating DOS and PDOS [http://abacus.deepmodeling.com/en/v3.7.3/advanced/elec_properties/dos.html]
• 63. What's new in Band 2025.1 - SCM [https://www.scm.com/doc/BAND/General/New_in_BAND.html]
• 64. Density of states [https://en.wikipedia.org/wiki/Density_of_states]
• 65. 1.3. Density of states calculation — RESCU 2.8.1 ... [https://docs.nanoacademic.com/rescu/tutorials/si_scf_dos_relax/dos/]
• 66. -Fold K - Cross Complete Guide [How To In ML & DL] validation [https://spotintelligence.com/2023/07/29/k-fold-cross-validation/]
• 67. Computational Methods in Drug Discovery - PMC [https://pmc.ncbi.nlm.nih.gov/articles/PMC3880464/]
• 68. Structure and dynamics in drug discovery [https://www.nature.com/articles/s44386-024-00001-2]
• 69. Combining K-fold cross validation with bayesian ... [https://www.nature.com/articles/s41598-025-23336-w]
• 70. Integrating over all orbitals - VASP Wiki [https://vasp.at/wiki/Integrating_over_all_orbitals]
• 71. A Recursive Hybrid Tetrahedron Method for Brillouin-zone ... [https://arxiv.org/html/2411.17162v1]
• 72. Comparison of the Tetrahedron Method to Smearing ... [https://arxiv.org/abs/2103.03469]
• 73. Beginner course @ CMD45 - the STATE hands-on tutorial! [https://state-tutorial.readthedocs.io/en/latest/CMD/CMD45/Beginner/index.html]
• 74. Band-structure calculation using hybrid functionals [https://vasp.at/wiki/Band-structure_calculation_using_hybrid_functionals]
• 75. Troubleshooting — BAND 2025.1 documentation - SCM [https://www.scm.com/doc/BAND/Troubleshooting/Trouble_Shooting.html]
• 76. inconsistent behavior between band structure and density ... [https://mattermodeling.stackexchange.com/questions/4096/inconsistent-behavior-between-band-structure-and-density-of-states]
• 77. Meta-GGA and 2D confined InAs - QuantumATK [https://docs.quantumatk.com/tutorials/inas_2d_mgga/inas_2d_mgga.html]