Atomic Orbital Basis Functions

Introduction

A molecular orbital is a spatial function, so a calculation must choose how to represent its shape. Most atom-centered electronic-structure calculations write each molecular orbital as a weighted sum of predetermined basis functions: ψᵢ(r) = Σμ Cμᵢχμ(r). The coefficient Cμᵢ is optimized by the calculation, while the form and location of each χμ come from the selected basis. This choice turns a continuous problem into a finite matrix problem. It is also an approximation: even a perfectly converged self-consistent-field iteration can only form orbitals from the shapes the basis permits.

Core explanation

Atomic orbitals are a natural starting point because molecules contain nuclei and much electron density remains concentrated around them. An atom-centered function has angular character such as s, p or d and a radial shape controlled by an exponent. Its name does not mean that it is an exact orbital of an isolated atom. In a molecule, the optimized molecular orbital is assembled from functions centered on one or several nuclei. The coefficients can change with geometry, charge and method, but fixed basis exponents generally do not change during a standard calculation. An orbital coefficient by itself is therefore not an observed atomic population.

For a simple s-like Slater-type orbital, the radial dependence is proportional to exp(−ζr), where r is distance from a nucleus and ζ controls spatial extent. Its exponential tail resembles the long-distance behavior of a bound atomic electron, and an s-type Slater function can have the appropriate cusp at a point nucleus. A larger ζ concentrates the function closer to the nucleus; a smaller ζ makes it more diffuse. More general Slater functions include a power of r and an angular factor. Their physically attractive radial form comes with difficult multicenter integrals when many electrons and several nuclei are involved.

A primitive Gaussian-type function instead has a radial factor such as exp(−αr²), multiplied when needed by Cartesian powers or angular functions. Because r² enters the exponent, a single s Gaussian has zero radial slope at the nucleus and decays too rapidly at large distance compared with a simple Slater exponential. That makes one primitive a poor stand-in for every part of an atomic orbital. The computational advantage is substantial: products of two Gaussians on different centers can be expressed through a Gaussian product center, making the many integrals needed by common quantum-chemical methods tractable. Washington State University's introductory quantum-chemistry notes discuss these radial forms and the distinction between primitives and their combinations.

Several Gaussian primitives with different exponents can be combined into a contracted basis function, χ = Σp dₚgₚ. Tight primitives describe density near a nucleus; more diffuse primitives help describe the outer region. The contraction coefficients dₚ are normally fixed by the basis definition, whereas molecular-orbital coefficients Cμᵢ are optimized in the molecular calculation. This is a nested expansion, not two independent sets of molecular orbitals. For example, the name STO-3G means that one Slater-like basis function is approximated by a contraction of three Gaussian primitives. It does not mean three independently adjustable valence functions for each atomic orbital. NIST's computational-thermochemistry glossary distinguishes contracted Gaussian-type orbitals and basis-family labels.

Angular flexibility and radial flexibility must be considered separately. Adding another s function with a different exponent improves radial freedom but cannot by itself bend an s-shaped density toward a bond. Adding p or d functions supplies additional angular patterns. Likewise, a very tight basis optimized for neutral compact atoms may poorly describe an anion's extended outer density. Subsequent pages distinguish split-valence, polarization and diffuse functions precisely because these changes solve different representational problems.

Basis functions are a representation choice, not a substitute for an electronic-structure method. Hartree–Fock with a large basis still omits correlation beyond its single determinant. A correlated method with a tiny basis may represent correlation but remain unable to express the needed orbital shapes. When a computed energy changes on enlarging a basis, that is evidence of basis incompleteness at the chosen method, not proof that all method error has vanished. NIST's basis-function listing makes the number of functions in named sets explicit, underscoring that a basis label specifies a finite mathematical space.

Step-by-step reasoning

1. Identify the nuclei and choose an atom-centered set with suitable radial and angular functions. 2. Distinguish each primitive Gaussian from the contracted function constructed from several primitives. 3. Expand each molecular orbital in the contracted functions and optimize its orbital coefficients. 4. Check whether the available radial and angular shapes can express bonding, lone pairs and extended density. 5. Repeat an important result with a systematically more flexible basis when accuracy matters. 6. Keep basis error separate from the electronic method's correlation or functional error.

Visual explanation

Sketch distance from a nucleus on the horizontal axis and orbital amplitude on the vertical axis. Draw a Slater-like exponential with a sharp initial slope and a persistent tail, then a single Gaussian with a flat initial slope and faster distant decay. Overlay a weighted sum of several Gaussians that follows the Slater-like curve more closely. In a second panel, place two nuclei under overlapping atomic-centered functions and show a molecular orbital as their weighted sum. Label fixed primitive and contraction coefficients separately from the optimized molecular-orbital coefficients.

Real-world analogy

Imagine constructing a curved outline from a set of reusable shaped tiles. The chosen tiles limit which contours are possible; rearranging their weights can improve a fit only within that collection. Contracting several Gaussian primitives is like permanently joining a few small tiles into one useful larger tile. This analogy explains finite representational freedom but not quantum interference, electron antisymmetry or the physical meaning of orbital phase.

Real-world example

A researcher estimates the electron affinity of a small molecule. A compact Gaussian basis gives a poorly bound added electron because its outer functions are too localized. The researcher adds diffuse functions and finds the electron affinity changes substantially. That sensitivity flags a representation problem; it does not by itself establish the experimental affinity, because method error, geometry, zero-point corrections and the treatment of the open-shell state remain. Reporting only the original basis label would hide why the first prediction was fragile.

Why?

Why use Gaussian functions if a Slater radial function better resembles an atomic orbital? A calculation repeatedly evaluates many one- and two-electron integrals over functions on different centers. Gaussian products have mathematical properties that make these integrals efficient. Several Gaussians can then be combined to recover much of the desired near-nuclear and outer-region shape. The tradeoff is that one Gaussian primitive is less physically shaped, so accuracy requires suitable contractions and extra functions for the chemistry being modeled.

Common misconception

“An atomic orbital basis function is an actual electron orbit.” It is a mathematical building block, while the calculated molecular orbitals are combinations of those blocks and neither is a classical trajectory. Another error is saying STO-3G provides three independently adjustable orbitals wherever a Slater orbital is expected: its three primitives are contracted into one fixed function in that role. A third is assuming that a converged SCF energy proves the basis is adequate. Numerical convergence only solves the chosen finite-basis equations.

Worked example

Suppose a hypothetical molecular orbital is expanded in two normalized atom-centered functions as ψ = 0.80χA + 0.60χB. If the functions were orthogonal, 0.80² + 0.60² = 1.00 would normalize this simple example. In an actual molecular calculation χA and χB usually overlap, so normalization instead requires CᵀSC = 1, where S contains overlaps. Now suppose χA itself is contracted from three primitives: χA = d₁g₁ + d₂g₂ + d₃g₃. The three d values are fixed basis coefficients, while 0.80 can change as the molecular orbital is optimized. Counting χA as three independent orbital basis functions would overstate the calculation's variational freedom. The example illustrates bookkeeping; the displayed values do not describe a real molecule.

Quick check

1. Why does a single Gaussian primitive poorly reproduce both the nuclear region and distant tail of a simple Slater function? Answer: Its radial slope at the center and its squared-distance exponential decay differ from the Slater function's cusp-like slope and exponential tail. 2. Which coefficients normally change during a standard molecular-orbital optimization, contraction coefficients or molecular-orbital coefficients? Answer: Molecular-orbital coefficients are optimized; the basis contraction coefficients are normally fixed by the selected basis definition.

Exam focus

Write a molecular orbital as a linear combination of atom-centered basis functions. Distinguish the exponential-in-r shape of Slater-type functions from the exponential-in-r² shape of Gaussian primitives. Explain why a contracted Gaussian uses multiple primitives yet behaves as one basis function for orbital expansion. Recognize that basis quality and electronic correlation are separate limitations. If asked about STO-3G, identify its three primitives as a contraction rather than a triple-zeta valence description.

Advanced insight

The exact complete-basis limit is a conceptual target, not a unique finite list of atom-centered functions. Different families can approach it at different rates for different observables. Tight functions matter for core-region properties, diffuse functions for weakly bound electrons, and polarization functions for anisotropic response. Even the optimal choice can depend on whether nuclei are represented by point charges or effective core potentials. A useful convergence study therefore compares the property of interest across sensible basis choices rather than declaring adequacy from total-energy stability alone.

Summary

Atomic-centered basis functions provide a finite vocabulary for molecular orbitals. Slater-type functions have useful exponential radial behavior; Gaussian primitives enable efficient multicenter integral evaluation. Contractions combine several primitives into a fixed function, after which molecular-orbital coefficients are optimized. A finite basis limits spatial flexibility independently of the electronic method's treatment of correlation. Basis selection should follow the property and electronic state being studied.

Practice questions

1. In ψᵢ = Σμ Cμᵢχμ, what does χμ represent? Answer: It is a selected basis function; the calculation adjusts its orbital coefficient Cμᵢ to build molecular orbital i. 2. What does the 3 in STO-3G count? Answer: Three Gaussian primitives contracted to approximate each Slater-like basis function in that minimal-basis construction. 3. Would increasing the number of Gaussian functions automatically add electron correlation to Hartree–Fock? Answer: No. A larger basis improves orbital representation, while Hartree–Fock still uses a single-determinant mean-field treatment. 4. Why might a diffuse basis function matter for an anion? Answer: An added electron may occupy spatially extended density that compact atom-centered functions cannot represent well.