How to Interpret TD-DFT Excitation Energies and Transitions

Time-dependent density functional theory (TD-DFT) calculations often print a list of excited states. Each entry may include an energy, a wavelength, an oscillator strength, and contributions from occupied-to-virtual orbital transitions. Those numbers answer different questions: energy locates a transition, oscillator strength estimates its optical intensity, and the transition description helps explain what changes in the electron distribution.

The interpretation below assumes a conventional calculation of singlet excitations from a closed-shell ground state. Open-shell systems, triplet calculations, and specialized methods require additional care.

What an excited-state entry means

A line such as “Excited State 1: Singlet-A 3.10 eV 400 nm f=0.002” describes a calculated transition from the ground state to the first excited state of the specified spin and symmetry, if those labels are printed. The energy is usually a vertical excitation energy: the energy difference evaluated at the ground-state nuclear geometry, without allowing the nuclei to relax in the excited state.

The wavelength is obtained from the energy using

$$\lambda\;(\mathrm{nm}) \approx \frac{1239.84}{E\;(\mathrm{eV})}.$$

The oscillator strength, $f$, is a dimensionless measure related to the strength of an electric-dipole transition. A larger value generally means a stronger absorption band, all else equal. A transition may have very small or zero oscillator strength because of symmetry or other selection rules, even though the excited state exists.

Reading the transition character

Many programs list contributions such as “HOMO → LUMO” or “HOMO−1 → LUMO.” These indicate which occupied and unoccupied molecular orbitals contribute to the excitation. They can give a quick qualitative picture, but a state is often a mixture of several orbital transitions. The printed coefficients depend on the program’s conventions and should not automatically be read as percentages or probabilities.

For a clearer picture, inspect natural transition orbitals (NTOs) or a transition-density plot if your software can produce them. NTOs compress a complicated transition into paired “hole” and “electron” orbitals: the hole shows where electron density is removed, and the electron orbital shows where it is added. This is often more informative than assigning a state to one HOMO–LUMO pair.

Look for whether the hole and electron occupy similar regions (often described as a local or valence excitation) or are spatially separated (a possible charge-transfer excitation). This distinction matters because some functionals, especially conventional ones with too little long-range exact exchange, can predict charge-transfer excitation energies poorly. A range-separated functional may be worth testing when substantial charge transfer is present.

Worked example: choosing a likely absorption band

Suppose a calculation reports these two singlet states:

State	Energy	Wavelength	Oscillator strength
1	3.10 eV	400 nm	0.002
2	2.60 eV	477 nm	0.35

The wavelengths follow from the energy conversion. For State 1, $1239.84/3.10 \approx 400$ nm; for State 2, $1239.84/2.60 \approx 477$ nm. State 2 is lower in energy and has a much larger oscillator strength, so it is expected to contribute more strongly to the calculated absorption spectrum. State 1 is much weaker, though it should not be discarded if weak features or state-specific chemistry matter.

Next, inspect the transition description or NTOs for both states. If State 2 is mainly a local excitation, its energy may be interpreted differently from a state with electron and hole density far apart. The table alone does not establish that character. Nor does it predict the exact experimental peak position: solvent, geometry, vibronic structure, temperature, and the computational method can all affect the comparison.

From a list of states to a spectrum

A vertical-excitation list is not automatically a simulated spectrum. To make one, broaden each transition using a chosen line shape, such as a Gaussian or Lorentzian, and combine the contributions. The chosen broadening width affects peak shape and overlap; it is not an additional TD-DFT prediction unless justified by a model.

When comparing with experiment, check that the conditions match as closely as practical. A gas-phase calculation and a solution-phase spectrum can differ because of solvent effects. Also distinguish an absorption spectrum from fluorescence: fluorescence generally involves relaxation on the excited-state surface before emission, so it is not simply the reverse of a vertical absorption calculation.

Common mistakes

• Calling every excitation a HOMO–LUMO transition. Several orbital pairs may contribute; inspect their relative contributions or use NTOs.
• Treating oscillator strength as a probability. It indicates relative optical intensity, not the chance that a molecule occupies that state under arbitrary conditions.
• Assuming the first state gives the strongest peak. State order is set by energy, while intensity depends on oscillator strength. A higher-energy state can absorb more strongly.
• Reporting too many digits. A computed energy with several decimal places does not imply that the method predicts an experimental band to that precision.
• Ignoring the method’s limitations. Standard linear-response TD-DFT can struggle with some charge-transfer, double-excitation, and near-degenerate states. Check whether the states of interest are plausible for the chosen functional and basis set.
• Comparing unlike quantities. Vertical absorption energies, relaxed excited-state energies, and emission energies refer to different geometries or processes.

A practical checklist

Before assigning a calculated band, record the ground-state geometry and calculation settings, confirm the state’s energy and oscillator strength, and inspect its transition density or NTOs. Then check whether the state has problematic charge-transfer or multiconfigurational character, and whether the calculated range of states covers the experimental region. For an important assignment, compare more than one suitable functional or computational approach rather than relying on a single orbital label.