Written and maintained by CASRAI Editorial Board
Published
On this page
A basis set is the finite collection of mathematical functions, centred on the atoms of a molecule, that a quantum-chemistry program uses to build the molecular orbitals in a calculation. The set you pick fixes how many functions each atom receives and of what shape (core, valence, polarization, diffuse), and it is part of the method: a result is only reproducible if the exact basis set, and a few related settings, are reported with it.
Short answer
In ab initio and density-functional calculations, orbitals are expressed as combinations of predefined functions, almost always Gaussian-type functions. The predefined collection is the basis set. Names such as STO-3G, 6-31G*, cc-pVTZ or def2-TZVP are compact descriptions of that collection. Because two programs or two papers can use the same family name for slightly different sets, the name, the function type (spherical or Cartesian), and any auxiliary or decontracted variants all belong in the methods section of a computational protocol.
Definition and scope
A minimal basis is described in the UC Davis Chem 110B course notes on extended basis sets as the smallest possible number of orbitals needed to calculate the electronic structure of a molecule: each atom contributes its occupied orbitals, usually the full partially filled shell. Everything larger is built by adding flexibility to that minimum.
Most practical sets use contracted Gaussian functions. A contracted function is one orbital function built from several Gaussian primitives whose exponents and coefficients are fixed in advance; during the calculation only the weights of the resulting functions vary. This is why a basis-set name encodes primitive counts as well as function counts. The oldest widely taught example, the STO-nG family, approximates each Slater-type orbital with n Gaussian-type orbitals, so STO-3G uses three primitives per function.
Basis-set choice is a modelling decision that sits next to the method (for example Hartree-Fock, density-functional theory or a correlated wavefunction method). It is not the same thing as the method, and it is also not the same thing as the software, although the program determines which sets are available and how they are stored. For a worked comparison of research software that runs such calculations, see the CASRAI guide to ORCA for quantum chemistry, and for a molecular-mechanics counterpart where no basis set is needed, molecular dynamics simulation.
How a basis set is built up in practice
Split-valence: more than one function per valence orbital
A minimal basis gives each orbital exactly one function. A split-valence basis keeps one function for each core orbital but uses multiple functions for each valence orbital, letting the calculation adjust how far valence electron density extends. This is the structure behind the Pople sets. In a Pople name of the form X-YG, X is the number of primitives in each core function and the digits of Y give the number of functions used for each valence orbital, each digit being that function’s primitive count. In 3-21G, the core function is three primitives and each valence orbital is represented by two functions, one of two primitives and one of one primitive. A name with three digits after the hyphen, as in 6-311G, therefore describes a triple-split valence.
Polarization functions
Polarization functions have higher angular momentum than the valence orbitals of the atom and allow electron density to deform in response to bonding. In the LibreTexts physical chemistry text on orbital polarization terms, a single star in a Pople name means functions of angular momentum one higher are added to second-row and heavier atoms (for example d functions on carbon), and a double star means hydrogen and helium also receive p functions. 6-31G** is synonymous with 6-31G(d,p) in that notation system, where the letters in brackets name the added functions explicitly.
Diffuse functions
Diffuse functions are very shallow Gaussians that describe the tail of an atomic orbital, far from the nucleus. In Pople names a single plus adds one diffuse function per valence orbital on all atoms except hydrogen and helium, and two plus signs (as in 6-311++G) also add them to hydrogen and helium. In Dunning sets the same idea is written as the prefix aug-. The Psi4 manual states that diffuse functions are generally necessary for anions, excited states and noncovalent interactions, which is the practical reason an anion study that uses an unaugmented set is a methodological red flag.
Correlation-consistent, Ahlrichs and Jensen families
The Psi4 manual lists its built-in sets by family. The Dunning correlation-consistent sets are named cc-pVXZ, where X runs through D, T, Q, 5 and 6 (the zeta level); in the Psi4 tables the highest angular-momentum function is d for D, f for T and g for Q. Variants add an extra d function for second-row p-block elements (cc-pV(X+d)Z), core-valence functions (cc-pCVXZ and cc-pwCVXZ), diffuse functions (aug- and its reduced forms), relativistic variants (the -DK suffix) and sets designed for explicitly correlated F12 methods. The Ahlrichs def2 family uses names such as def2-SVP, def2-TZVP and def2-QZVP, and the Jensen pcseg-n family is numbered so that n corresponds to n+1 zeta. Psi4 also notes that heavy-aug-cc-pVXZ and jul-cc-pVXZ are identical sets that carry two names.
The data table below the guide lists these naming conventions side by side with what each part of the name tells you.
Reading and writing basis-set names: a notation decoder
When you read a methods section, parse the name from left to right. The family prefix (Pople digits, cc-, def2-, pcseg-) tells you which design philosophy the set follows. The zeta label (D, T, Q, SVP, TZVP, the digits of Y) tells you how many functions are given to valence orbitals. The decorations tell you what was added to that skeleton: stars or letters in brackets for polarization, plus signs or aug- for diffuse functions, C or wC for core correlation, -DK for a relativistic recontraction, -RI or similar suffixes for auxiliary fitting sets. If any one of those cannot be read from the paper, the calculation is not fully specified.
A worked reading: 6-311++G** is a Pople set with 6 primitives in the core function and a triple-split valence of three functions per valence orbital, with diffuse functions on every atom including hydrogen and helium (the two plus signs) and polarization functions on every atom including hydrogen and helium (the two stars). Another: aug-cc-pVTZ is a Dunning correlation-consistent set at triple-zeta level (highest angular momentum f in the Psi4 tables) with diffuse functions added through the aug prefix.
What to record in a methods section or protocol
The Psi4 manual and the Basis Set Exchange project together show what a basis set actually involves beyond its headline name. The following checklist is drawn from those two sources and is meant as a reproducibility aid for computational protocols; pair it with the guidance in writing the methods section so it can be reproduced.
- The exact name. Use the name the program uses, not a shortened family label. In Psi4 the orbital basis is selected with
set basis NAME. - Per-atom assignments. Psi4 can assign different sets to different atoms inside a
basis { assign ... }block (for example one set on all carbons and another on a labelled hydrogen). A single name in the text is not enough if the input file uses assignments. - Spherical or Cartesian functions. Psi4 supports both spherical harmonic (5D/7F) and Cartesian (6D/10F) Gaussians but not a mixture, either within a basis set or within one calculation. For built-in sets the setting follows the orbital basis; for user-defined sets it comes from the first line of the file. Results from programs with different defaults can therefore differ even when the family name matches.
- Auxiliary (fitting) sets. Density-fitted methods use a second set, set separately in Psi4 (for example
set df_basis_mp2 cc-pvdz-ri). Report it. - Decontraction. Appending
-deconto a name decontracts the set. The Psi4 manual notes this adds significant computational cost and recommends decontracted sets for relativistic X2C calculations. A decontracted run is a different basis set and must be labelled as such. - Source and version of the set. The Basis Set Exchange, run by MolSSI and EMSL, distributes sets through a web front end, a Python library and command-line tool and REST APIs, and exports references in BibTeX and JSON. Original sets are kept as version 0 and updated sets verified against authoritative sources are version 1, so state which version you downloaded and cite the sets themselves as well as the exchange.
Where it matters for labs and research groups
Computational chemistry groups share inputs and outputs among students, collaborators and reviewers, so a basis set defined only by a family name is a recurring source of failed replication. Two practical consequences follow from the sources above.
First, file-system and software conventions can break a name. The Psi4 manual stores 6-31++G** in a file called 6-31ppgss.gbs, because the plus sign becomes pp and the star becomes ss in lower case. A script that builds filenames from basis-set labels, or a workflow manager that archives input decks, needs to account for this. Version control for such inputs is covered in the CASRAI guide to computational reproducibility.
Second, sets are tools for a specific job. The Psi4 tables show that families exist for core-valence correlation, relativistic treatment, and explicitly correlated methods. Choosing a plain valence set for a property that depends on core electrons, or omitting diffuse functions for an anion, is a modelling error that reviewers can detect from the name alone. A short sentence in the protocol justifying the family and the decorations is cheap and prevents that critique.
Data-management plans for computational projects should likewise list the basis-set files or the exchange version alongside the input decks and outputs; this lets a later reader regenerate the calculation without guessing.
Common misunderstandings
“The basis set is just a setting in the software”
It is part of the model. The name, the function type, auxiliary sets and decontraction all change the set of functions the orbitals are built from, so each is part of the method description.
“The same name means the same functions everywhere”
Not necessarily. The Basis Set Exchange keeps original sets as version 0 and verified updates as version 1, and individual programs store and interpret sets with their own conventions (spherical versus Cartesian, file naming). Record the source and version rather than assuming equivalence.
“A star and a plus mean the same kind of addition”
They do not. Stars (or letters in brackets) add polarization functions of higher angular momentum that let density deform; plus signs add diffuse functions that extend the orbital tail. They answer different physical needs, which is why anions and noncovalent interactions call for diffuse functions while bonding geometry calls for polarization.
“Two names mean two different sets”
Sometimes one set carries two names. The Psi4 manual lists heavy-aug-cc-pVXZ and jul-cc-pVXZ as identical, and 6-31G** and 6-31G(d,p) are synonymous in Pople notation.
“A bigger name always means a different family”
Zeta labels follow different conventions in different families: Pople names count primitives in digits, Dunning names use D, T, Q, 5 and 6, def2 names use SVP, TZVP and QZVP, and Jensen pcseg-n numbers are offset so that n corresponds to n+1 zeta. Comparing “n” across families without checking the convention is a common slip.
Frequently asked questions
What is a basis set in computational chemistry?
It is the predefined collection of functions, usually contracted Gaussians centred on the atoms, from which a program builds molecular orbitals. Its name encodes how many functions each orbital gets and which polarization or diffuse functions are added.
What does 6-31G* mean?
The 6 gives the number of primitives in each core function, and 31 describes a split valence with two functions per valence orbital of three and one primitives. The star adds polarization functions (d on second-row and heavier atoms). 6-31G** also adds p functions on hydrogen and helium and is the same as 6-31G(d,p).
When are diffuse functions needed?
The Psi4 manual says they are generally necessary for anions, excited states and noncovalent interactions. Add them through plus signs in Pople names or the aug- prefix in Dunning names.
Can I mix spherical and Cartesian functions in one calculation?
Not in Psi4: both are supported (5D/7F and 6D/10F) but mixtures are not, within a basis set or within a calculation.
Where can I download basis-set definitions?
The Basis Set Exchange from MolSSI and EMSL provides a web frontend, a Python library and command-line tool, and REST APIs, with references exported as BibTeX or JSON.
What should I cite for a basis set?
Cite the original publication of the set and the exchange paper by Pritchard and co-authors (J. Chem. Inf. Model. 59(11), 4814-4820, 2019, doi 10.1021/acs.jcim.9b00725); the exchange supplies reference lists for each set.
Sources
- UC Davis Chem 110B, LibreTexts, section 3.04 Extended Basis Sets: chem.libretexts.org, read 2026-11-15.
- LibreTexts Physical Chemistry, 11.4 Orbital Polarization Terms in Basis Sets: chem.libretexts.org, read 2026-11-15.
- Psi4 manual, Basis Sets: psicode.org/psi4manual/master/basissets.html, read 2026-11-15.
- Psi4 manual, basis sets by family: psicode.org/psi4manual/master/basissets_byfamily.html, read 2026-11-15.
- MolSSI, announcement of the new Basis Set Exchange: molssi.org, read 2026-11-15; Pritchard et al., J. Chem. Inf. Model. 59(11):4814-4820 (2019), doi:10.1021/acs.jcim.9b00725.
Data: Basis-set naming conventions: what each family and suffix encodes
Compiled from a university physical-chemistry course text, the LibreTexts orbital-polarization chapter and the Psi4 manual's basis-set pages: for each commonly seen basis-set name or decoration, the family it belongs to and what the name tells a reader about the set.
Period covered: 2026.
| Name or suffix | Family | What the name encodes |
|---|---|---|
| STO-3G | Pople (STO-nG) | Each Slater-type orbital approximated by 3 Gaussian primitives; a minimal basis |
| 3-21G | Pople split-valence | Core function of 3 primitives; each valence orbital as two functions of 2 and 1 primitives |
| 6-31G* | Pople | 6-primitive core; two valence functions of 3 and 1 primitives; polarization functions on second-row and heavier atoms |
| 6-31G** (= 6-31G(d,p)) | Pople | As 6-31G* plus p polarization functions on hydrogen and helium |
| 6-311++G | Pople | Triple-split valence; diffuse functions on all atoms including hydrogen and helium |
| cc-pVXZ (X = D, T, Q, 5, 6) | Dunning correlation-consistent | Zeta level X; highest angular momentum d (D), f (T), g (Q) |
| cc-pV(X+d)Z | Dunning | Extra d function for second-row p-block elements |
| cc-pCVXZ, cc-pwCVXZ | Dunning core-valence | Added functions for core and core-valence correlation |
| aug- prefix | Dunning | Diffuse functions added; needed for anions, excited states and noncovalent interactions |
| -DK suffix | Dunning | Relativistic variant of the Dunning set |
| def2-SVP, def2-TZVP, def2-QZVP | Ahlrichs def2 | Valence split and zeta level named in the suffix: SVP, TZVP, QZVP |
| pcseg-n | Jensen | Number n corresponds to n+1 zeta |
| -decon suffix (Psi4) | Software convention | Decontracts the named set; adds significant computational cost |
Sources
- UC Davis Chem 110B, 3.04 Extended Basis Sets (LibreTexts) (accessed 2026-11-15)
- LibreTexts Physical Chemistry, 11.4 Orbital Polarization Terms in Basis Sets (accessed 2026-11-15)
- Psi4 manual: Basis Sets (accessed 2026-11-15)
- Psi4 manual: basis sets by family (accessed 2026-11-15)
Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.







