A Byte of PhysicsLogo

Computational materials

Computational materials science connects atomic structure to predicted energies, forces, geometries, electronic bands, vibrations, and response properties. Each result belongs to a particular physical approximation and finite numerical representation. A band gap or formation energy without that context is not a reproducible material claim.

Calculations range from empirical interatomic potentials to quantum electronic-structure methods. The right method depends on the observable, chemical environment, system size, temperature, and accuracy needed.

Separate nuclei and electrons carefully

Many electronic-structure calculations use the Born–Oppenheimer approximation: nuclei are treated as fixed while the electronic state is solved, then nuclear positions are updated using the resulting forces. This is useful because nuclei are much heavier than electrons, but it can fail when electronic and nuclear motion are strongly coupled.

For fixed nuclear positions \(\{\mathbf R_I\}\), the electronic solution defines a potential-energy surface \(E(\{\mathbf R_I\})\). Forces are

\[\mathbf F_I=-\frac{\partial E}{\partial\mathbf R_I}\]

at a fully self-consistent electronic state, with additional implementation details such as Pulay corrections when the basis depends on geometry.

Density-functional theory

The many-electron wavefunction depends on three coordinates per electron and becomes extremely expensive as electron count grows. Density-functional theory uses the electron density

\(n(\mathbf r)\)

as its central variable. In Kohn–Sham DFT, auxiliary one-electron orbitals obey

\[\left[-\frac{\hbar^2}{2m_e}\nabla^2+V_{\mathrm{eff}}[n](\mathbf r)\right]\phi_i(\mathbf r)=\epsilon_i\phi_i(\mathbf r)\]

and the density is constructed from occupied orbitals:

\[n(\mathbf r)=\sum_i f_i|\phi_i(\mathbf r)|^2\]

The effective potential includes nuclear attraction, classical electron repulsion, and an exchange-correlation approximation. The exact exchange-correlation functional is unknown, so choices such as local, generalized-gradient, hybrid, or more specialized approximations create model error even after numerical convergence.

Kohn–Sham eigenvalues are useful computational quantities, but they are not all automatically exact excitation energies. Standard approximations often underestimate semiconductor band gaps.

The self-consistent field loop

Because the potential depends on density and the density depends on the orbitals, the equations are solved iteratively:

import numpy as np

# Two-site mean-field toy model illustrating self-consistency, not a DFT calculation.
density = np.array([0.7, 0.3])
for iteration in range(1000):
    H = np.array([[density[0], -1.0], [-1.0, density[1]]])
    energies, orbitals = np.linalg.eigh(H)
    output_density = abs(orbitals[:, 0]) ** 2
    residual = np.linalg.norm(output_density - density)
    density = 0.7 * density + 0.3 * output_density
    if residual < 1e-10:
        break
else:
    raise RuntimeError("Self-consistency failed to converge.")
print("Converged site populations:", density, "residual:", residual)

Simple replacement can oscillate or diverge, especially in metals. Mixing and preconditioning improve convergence. Reaching the iteration threshold means the chosen discrete equations are self-consistent; it does not establish that the functional, structure, or temperature is physically adequate.

Basis sets and pseudopotentials

Electronic orbitals need a finite basis. Plane waves are systematic for periodic materials and are controlled by an energy cutoff:

\[\frac{\hbar^2|\mathbf G+\mathbf k|^2}{2m_e}\leq E_{\mathrm{cut}}\]

Increasing \(E_{\mathrm{cut}}\) adds basis functions and usually increases cost. Localized atomic orbitals can be more compact but require their own size and completeness tests.

Core electrons oscillate rapidly near nuclei and are often replaced by pseudopotentials or projector-augmented-wave data. The chosen data set must match the exchange-correlation approximation and include relevant semicore, relativistic, and spin–orbit effects. Two calculations with the same cutoff but different pseudopotentials are not the same model.

Reciprocal-space sampling

Periodic crystals require integration over the Brillouin zone. A finite \(\mathbf k\)-point mesh approximates that integral:

\[\frac{1}{\Omega_{\mathrm{BZ}}}\int_{\mathrm{BZ}}Q(\mathbf k)\,d\mathbf k\approx\sum_jw_jQ(\mathbf k_j)\]

Metals often need denser sampling than wide-gap insulators because occupation changes sharply near the Fermi surface. Smearing can aid integration, but its width affects energies and forces and must be converged or extrapolated appropriately.

# Brillouin-zone average for a fixed illustrative tight-binding dispersion.
previous = None
for points in [16, 32, 64, 128]:
    k = 2 * np.pi * (np.arange(points) + 0.5) / points - np.pi
    observable = np.mean(np.sqrt(1 + 0.5 * np.cos(k)))
    change = None if previous is None else abs(observable - previous)
    print(points, observable, change)
    previous = observable

Converge the actual target observable. Total energy per atom may appear stable before a small band gap, stress, phonon frequency, or magnetic energy difference does.

Periodic cells and finite-size effects

Periodic boundaries represent an infinite repeating structure. A primitive cell is efficient for a perfect crystal. Defects, surfaces, disorder, and isolated molecules need larger supercells plus checks that repeated images do not interact significantly.

A charged defect is especially delicate because periodic copies and compensating background charge produce long-range artifacts. Formation energies require finite-size and potential-alignment corrections whose definitions must be reported.

For a slab, vacuum thickness, slab thickness, surface termination, dipole corrections, and in-plane cell size all need convergence. “Twenty angstroms of vacuum” is an input, not proof of isolation.

Relaxing a structure

A geometry optimization seeks a local stationary structure. Typical conditions include small maximum force and small stress:

\[\max_I|\mathbf F_I|<F_{\mathrm{tol}},\qquad \max_{\alpha\beta}|\sigma_{\alpha\beta}-\sigma_{\alpha\beta}^{\mathrm{target}}|<\sigma_{\mathrm{tol}}\] 
# Relax a two-atom bond in a harmonic potential, in scaled units.
separation, equilibrium, stiffness = 1.5, 1.0, 4.0
for iteration in range(100):
    force = -stiffness * (separation - equilibrium)
    if abs(force) < 1e-9:
        break
    separation += 0.1 * force
print("Relaxed separation:", separation, "force:", force)

A small force identifies a stationary point, not necessarily the global minimum. Different starting structures can relax to different metastable configurations. Check symmetry, unstable vibration modes when relevant, and alternative magnetic or structural starting states.

Useful energy differences

Absolute total energies depend strongly on method details. Carefully constructed differences are more interpretable. A cohesive energy compares a crystal with isolated constituent atoms under consistent conventions. A formation energy compares a compound with reference chemical potentials:

\[\Delta E_{\mathrm f}=E_{\mathrm{compound}}-\sum_i n_i\mu_i\]

The reference phases and per-atom or per-formula-unit normalization must be specified. Comparing energies from incompatible pseudopotentials, functionals, spin states, or numerical settings can create meaningless differences.

For a neutral defect,

\[E_{\mathrm f}^{\mathrm{defect}}=E_{\mathrm{defective}}-E_{\mathrm{perfect}}-\sum_i\Delta n_i\mu_i\]

Sign conventions for atoms added or removed must be declared. Charged defects also depend on electron chemical potential and electrostatic corrections.

Mechanical and vibrational properties

Elastic constants can be obtained from small strains and energy or stress responses:

\[E(V,\boldsymbol\epsilon)=E_0+\frac{V_0}{2}\sum_{ij}C_{ij}\epsilon_i\epsilon_j+\cdots\]

Strains must be small enough for linear elasticity but large enough to overcome numerical noise. Multiple positive and negative strains help reveal nonlinear contamination.

Phonon frequencies come from second derivatives of energy with respect to atomic displacements. Imaginary frequencies are commonly represented as negative plotted values and indicate instability of the reference structure within the harmonic approximation, not literal negative oscillation frequencies.

Zero-temperature electronic energies omit thermal expansion, vibrational free energy, configurational disorder, and anharmonic effects. Comparison with a room-temperature experiment needs those differences addressed.

Classical and learned interatomic models

Empirical force fields and machine-learned potentials can evaluate much larger systems and longer trajectories than first-principles methods. Their accuracy is restricted by functional form or training data. A small force error on a test set does not prove reliability for new phases, defects, charge states, fracture, or chemical reactions.

Validate energies, forces, stresses, symmetries, and target properties on structures separate from training. Detect extrapolation where possible. The reference electronic method also has model error, so a learned model cannot automatically exceed the truth of its labels.

A convergence matrix

Change numerical controls separately before combining uncertainties:

for stiffness in [3.0, 4.0, 5.0]:
    for tolerance in [1e-4, 1e-8]:
        separation = 1.5
        for _ in range(1000):
            force = -stiffness * (separation - 1.0)
            if abs(force) < tolerance:
                break
            separation += 0.1 * force
        print(stiffness, tolerance, separation, 0.5 * stiffness * (separation - 1) ** 2)

Numerical convergence and physical validation answer different questions. A result can be converged to many digits and still be wrong because the functional, structure, phase, or temperature is unsuitable.

Reporting a reusable result

Archive input and output structures, units, code and library versions, pseudopotential checksums, basis and cutoff, k-point mesh, occupations, charge, spin, convergence thresholds, symmetry handling, and post-processing commands. Report raw calculated quantities separately from corrections and experimental comparisons.

The introductory references explain atoms, bands, and condensed matter. The specialized Materials Project methodology documents a large-scale electronic-structure workflow. Neither makes settings transferable without convergence tests for the property and material at hand.

A worked Python notebook

Relax an illustrative bond toward its equilibrium length of one. The question for this calculation is precise: how does harmonic bond energy change when we change bond length, while keeping the other choices fixed? The earlier sections explain the physical relationship. Here we turn one part of that relationship into a calculation that you can run, inspect, and change in Google Colab.

This toy potential illustrates optimisation; it is not a density-functional calculation or a prediction for a named material. Keep that sentence beside the results. A formula can return a number outside its useful range, and a plotting library will usually draw that number without asking whether the physical assumptions still hold. The selected interval is from 0.7 to 1.5 scaled length units. It gives us a specific domain to discuss instead of leaving the graph's range to a software default.

Use the notebook link above the lesson, run the setup cell, and then run the cells in order. The code below is also included there. NumPy handles the arrays of numerical values, SymPy keeps the equation symbolic, and Matplotlib draws the results. If these libraries or notebook cells are unfamiliar, start with Your first physics notebook.

State the relationship before choosing a graph

In this small example, q stands for bond length, expressed in scaled length units. The output is harmonic bond energy, expressed in scaled energy units. The equation below uses the numerical values in those units. Any fixed coefficients therefore carry the units needed to make the relation consistent. A value supplied in another unit must be converted before it enters this expression.

\[f(q) = 2 \left(q - 1\right)^{2}\]

This is a function of one changing input. It is not a statement that the physical system has only one variable. We deliberately hold the other inputs fixed so that one cause of variation is visible. Changing several inputs at once would make it harder to tell which change produced the output difference. Write down the held-fixed choices before trying a different experiment.

At 1.2 scaled length units, the expression gives 0.08 scaled energy units. This is our reference calculation. It is a prediction of the stated example, not a new measurement. Its displayed digits help us compare implementations. They should not be read as a claim that physical inputs are known to six significant figures.

import numpy as np
import sympy as sp
import matplotlib.pyplot as plt

q = sp.Symbol("q", real=True)
expression = sp.sympify("2*(q-1)**2", locals={"q": q})
reference_input = 1.2
reference_output = float(expression.subs(q, reference_input))
print("Input:", reference_input, "scaled length units")
print("Predicted harmonic bond energy:", reference_output, "scaled energy units")

sp.Symbol declares a mathematical symbol. It does not assign a numerical measurement to that symbol. subs then replaces the symbol with our chosen input, and float converts the resulting symbolic number into a standard Python numerical value. Keeping these steps separate lets us inspect the equation before numerical substitution. It also prevents a later plot from becoming the only surviving record of the model.

Compare a few cases you can explain

Before drawing hundreds of points, calculate three. The lower endpoint gives 0.18 scaled energy units. The midpoint, at 1.1 scaled length units, gives 0.02 scaled energy units. The upper endpoint gives 0.5 scaled energy units. Read these together with the assumptions: they describe different choices of bond length within the same example, with the other inputs held fixed.

These three values are a useful first check, but they do not establish the shape between them. Two endpoints can be equal even when the curve bends or oscillates in between. Likewise, a large change between endpoints does not tell us whether the relationship is linear. We need the equation and intermediate samples to answer those questions.

evaluate = sp.lambdify(q, expression, modules="numpy")
inputs = np.array([0.7, 1.1, 1.5])
outputs = np.asarray(evaluate(inputs), dtype=float)
outputs = np.broadcast_to(outputs, inputs.shape)
for input_value, output_value in zip(inputs, outputs):
    print(
        f"{input_value:.6g} scaled length units -> "
        f"{output_value:.6g} scaled energy units"
    )
assert np.all(np.isfinite(outputs))

lambdify turns the symbolic expression into a numerical function. We explicitly select NumPy as its numerical backend so it can accept an array. The same expression now has two useful forms: one for algebra and one for calculating many cases. The assertion checks that the chosen inputs produce finite outputs. It would catch an infinity or an undefined numerical result, but a finite value can still be physically inappropriate.

Make a graph that answers the question

The horizontal axis shows bond length; the vertical axis shows harmonic bond energy. Every plotted point is a prediction of the same expression. We mark the reference case separately so that a reader can connect the earlier printed value to its place on the curve. The axis labels include units because a shape without a scale cannot tell us the size of a physical effect.

input_grid = np.linspace(0.7, 1.5, 301)
output_grid = np.asarray(evaluate(input_grid), dtype=float)
output_grid = np.broadcast_to(output_grid, input_grid.shape)
fig, ax = plt.subplots(figsize=(7, 4))
ax.plot(input_grid, output_grid, label="Stated model")
ax.scatter([reference_input], [reference_output], label="Reference case", zorder=3)
ax.set_xlabel("Bond length (scaled length units)")
ax.set_ylabel("Harmonic bond energy (scaled energy units)")
ax.legend()
ax.grid(alpha=0.25)
fig.tight_layout()
plt.show()

Look for a sign change, a turning point, or a region where the curve becomes steep. Connect each feature to the expression rather than to the colour of the line. If the curve has a minimum or maximum, changing the input in opposite directions may have similar effects on the output near that point. If it is nearly straight over a short interval, a local linear approximation may be useful even when the complete expression is nonlinear.

Increasing the number of samples makes the line smoother and may reveal features that a coarse grid missed. It does not improve the physical inputs or enlarge the model's validity range. For quantities that can take only discrete values, use the line as a guide to the relationship and interpret only the allowed input values. A plotting function does not enforce such restrictions for you.

Measure the local sensitivity

A derivative asks a more specific question than the whole graph: how quickly does harmonic bond energy change near the reference input? Its units are the output unit divided by the input unit. For this calculation, the derivative at 1.2 scaled length units is 0.8 in those ratio units. Locally, the output increases as the input increases. This statement is local; the sign or size of the slope can be different elsewhere on the curve.

\[f'(q) = 4 q - 4\]

For a small input change, multiply the derivative by that change to estimate the output change. Then compare that estimate with a fresh evaluation of the full expression. The distinction between these two calculations matters: the derivative gives a tangent approximation, while substitution evaluates the chosen expression at the new point.

slope_expression = sp.diff(expression, q)
slope_at_reference = float(slope_expression.subs(q, reference_input))
changed_input = 1.204
linear_prediction = reference_output + slope_at_reference * (
    changed_input - reference_input
)
full_prediction = float(evaluate(changed_input))
print("Symbolic derivative:", slope_expression)
print("Tangent prediction:", linear_prediction, "scaled energy units")
print("Full expression:", full_prediction, "scaled energy units")
print(
    "Approximation difference:",
    full_prediction - linear_prediction,
    "scaled energy units",
)

Here the changed input is 1.204 scaled length units. The tangent estimate is 0.0832 scaled energy units, while the full expression gives 0.083232 scaled energy units. If the two values are close, it means the chosen input change is small enough for this local approximation at this point. It does not mean the entire model is linear. If the slope is zero, a quadratic or higher-order term may supply the leading change instead.

This is also the beginning of uncertainty propagation. If the input has a small stated standard uncertainty, the magnitude of the slope times that uncertainty estimates one contribution to output uncertainty. Do not automatically use the same rule for a wide interval, a discontinuity, a turning point with zero first derivative, or several correlated uncertain inputs. Those situations need more of the model than one slope.

Check the derivative with numerical differences

We can estimate the same slope from values on either side of the reference point. Subtract the lower output from the upper output and divide by twice the input step. Parentheses are essential: the entire output difference must be divided by the entire input interval. Without them, Python can execute a different mathematical expression while still producing a plausible number.

The calculation below uses three decreasing step sizes. Each estimate is compared with the symbolic derivative, so the reference is independent of the numerical difference formula. Near an endpoint, it is acceptable here to evaluate a small extension only because the displayed mathematical expression is smooth there; such a check does not assert that extrapolated values describe additional physical cases.

input_span = 0.8
for fraction in [1e-2, 1e-3, 1e-4]:
    step = input_span * fraction
    upper = float(evaluate(reference_input + step))
    lower = float(evaluate(reference_input - step))
    numerical_slope = (upper - lower) / (2 * step)
    difference = numerical_slope - slope_at_reference
    print(f"step={step:.6g}: slope={numerical_slope:.6g}, difference={difference:.6g}")

For a smooth function, reducing the step often reduces the approximation error at first. Very small steps can eventually make subtraction of nearly equal floating-point values less reliable. Therefore “smaller” is not an unlimited guarantee of “better.” Compare the observed behaviour over a useful range of steps instead of choosing the smallest number the computer can represent.

An absolute slope difference is useful when the expected slope is near zero. A relative difference would divide by a very small reference and could look enormous even when the absolute discrepancy is tiny. Away from zero, a relative difference can help compare quantities with very different scales. The tolerance should reflect the question and the numerical method, rather than a universal decimal threshold copied into every test.

Change the example and preserve the reasoning

Make one copy of the reference calculation before editing it. Choose a second value of bond length inside the stated interval, predict whether harmonic bond energy will rise or fall, and then evaluate the expression. Place the prediction and the result in a text cell. If they disagree, identify whether the mistake was in the sign, the unit conversion, the algebra, or the assumption about how the curve behaves.

Next choose a shorter plotting interval around the reference point. Compare the local curve with its tangent approximation. Then widen the interval again. The difference shows why a useful local estimate need not describe the whole range. Keep the axis labels and numerical scale visible when comparing the plots; automatic axis limits can make two different effects appear equally large.

Finally, return to the physical limit stated at the start: This toy potential illustrates optimisation; it is not a density-functional calculation or a prediction for a named material. Describe one change to the setup that would violate that limit. State what new input, equation, or observation would be needed before trusting an answer in the changed situation. This last step connects the Python experiment to the physics instead of treating successful execution as the only criterion for a useful result.

Before saving, restart the runtime and run all cells in order. A result that depends on an earlier, now-deleted input cell is not reproducible from the saved notebook. Keep unrounded values for later calculations, round only the reported output, and label any new data as measured, assumed, or generated. The graph and printed values should remain traceable to the same expression and the same choices of inputs.

Recover the output change by integrating the slope

There is another way to check that the expression and its derivative describe the same relationship. Add up the small output changes predicted by the slope across the chosen input interval. In the continuous limit, this integral equals the output at the upper endpoint minus the output at the lower endpoint. The result has units of scaled energy units: the input unit in the integration cancels the input unit in the denominator of the slope.

This statement is about the mathematics of this particular relationship. It does not assign a new physical meaning to an arbitrary area under a graph. For example, an area under force versus distance represents work, while an area under a different pair of quantities may represent something else or have no standard name. Always carry the axis units through the calculation before naming an integral.

slope_function = sp.lambdify(q, slope_expression, modules="numpy")
endpoint_change = float(evaluate(1.5) - evaluate(0.7))
for points in [101, 201, 401]:
    grid = np.linspace(0.7, 1.5, points)
    slopes = np.asarray(slope_function(grid), dtype=float)
    slopes = np.broadcast_to(slopes, grid.shape)
    integrated_change = np.trapezoid(slopes, grid)
    print(
        points,
        "integrated change:",
        integrated_change,
        "endpoint change:",
        endpoint_change,
    )

The trapezoidal rule joins neighbouring slope samples with straight segments and adds their signed areas. A negative contribution remains negative; replacing every slope by its magnitude would calculate a different quantity. Increasing the sample count provides a practical refinement check. If the values do not settle toward the endpoint difference, first inspect the interval, singularities, and array ordering.

Do not conclude that two agreeing calculations prove the model is physically complete. Both use the same selected relationship, so an omitted effect would be omitted from both. Their agreement is evidence that the numerical tools implement that relationship consistently. Comparing with an independent experiment or a more appropriate physical model is a separate test.

Keep this comparison in the notebook when you change the expression. It is particularly useful after adding a coefficient, changing a sign convention, or converting an input unit. A code cell that produces a smooth graph can conceal these mistakes; a second calculation of the same output change gives you another place to notice them.

Sources

Share to: