Kramers–Kronig relations

A linear response function describes how a system reacts to an applied signal at different frequencies. It is generally complex-valued: its real and imaginary components encode complementary aspects of the response. For electromagnetic materials, these components are associated with dispersion and absorption, though the precise interpretation depends on the response function and sign convention used. The Kramers–Kronig relations are integral transforms relating the two components across the frequency spectrum; they do not equate the real and imaginary parts at one frequency alone.

Their central physical condition is causality: a material or system cannot respond before it is stimulated. In the frequency domain, causality implies analytic properties of the response function, subject to suitable decay and regularity conditions. Those properties make the real and imaginary components mathematically interdependent. The same broad principle applies to many causal linear-response settings, not just optical materials.

If the imaginary part of a response is known across a sufficiently broad frequency range, a dispersion relation can be used to calculate the corresponding real part, and conversely. In optics this connects absorption with refractive dispersion. In electron energy-loss spectroscopy, measured spectral information can be processed using Kramers–Kronig analysis to infer components of the complex dielectric function and thereby estimate optical properties. The method can reveal response features that are not obvious in the raw spectrum alone.

The relations are also useful as consistency checks. Independently measured real and imaginary response curves that violate the appropriate Kramers–Kronig relation may signal measurement problems, unaccounted background contributions, or an unsuitable model. Such a check depends on applying the correct relation and assumptions for the chosen physical quantity.

The integrals formally involve a broad, often unbounded, frequency range, whereas experiments cover only a finite interval. Analysts therefore need to estimate or model contributions outside the measured band. For example, optical-property estimates from reflectance data may require extrapolations at low and high frequencies. Different plausible extrapolations can change the result, particularly near the edges of the measured range.

Real datasets also require corrections for instrumental effects, noise, baseline offsets, and the specific measured quantity. A Kramers–Kronig transform does not remove these uncertainties; it propagates the input data and assumptions into the inferred component. Thus the result should be interpreted alongside the measurement range, preprocessing and extrapolation choices, rather than as an assumption-free reconstruction.

The relations are named for the physicists Ralph Kronig and Hendrik Anthony Kramers, whose work in the 1920s established the dispersion relations in the context of optical response and theoretical physics. They are closely related to Hilbert-transform pairs and are often called dispersion relations. Their enduring importance comes from a general consequence of causality in frequency-domain response, while their concrete formula and sign conventions depend on definitions and the physical system being analyzed.