# Introduction — What the Kramers–Kronig Relations Say *A first statement of the two integral relations, the physical principle they encode, and a roadmap for the series.* ## The two integral relations Let χ(ω) = χ′(ω) + iχ″(ω) be a complex response function of real angular frequency ω — for example the electric susceptibility of a dielectric, the complex refractive index minus one, a mechanical compliance, or the complex permeability of a magnetic material. Under the assumptions spelled out in later chapters (the system is linear, time-invariant, causal, and χ(ω) decays fast enough at high frequency), the real and imaginary parts of χ are **not independent**. They are locked to one another through the pair of Hilbert-transform integrals known as the Kramers–Kronig (K–K) relations: ``` χ′(ω) = (2/π) P ∫₀^∞ [ω′ χ″(ω′) / (ω′² − ω²)] dω′ χ″(ω) = −(2/π) P ∫₀^∞ [ω χ′(ω′) / (ω′² − ω²)] dω′ ``` Here `P` denotes a Cauchy principal value (Chapter 8), needed because the integrand has a pole at ω′ = ω on the real axis. The factors assume χ is the response of a **real-valued** input–output system, so χ(−ω) = χ*(ω), which lets the integration range be folded from (−∞, ∞) to (0, ∞). In the most general two-sided form (before folding), ``` χ′(ω) = (1/π) P ∫_{−∞}^{∞} χ″(ω′) / (ω′ − ω) dω′ χ″(ω) = −(1/π) P ∫_{−∞}^{∞} χ′(ω′) / (ω′ − ω) dω′ ``` which exhibits the structure more transparently: χ′ and χ″ are a **Hilbert-transform pair**. ## What the relations mean physically The two halves of χ describe different observable phenomena: - **χ′(ω)** — the *reactive*, *dispersive*, in-phase part: refractive index, phase velocity, elastic compliance, stored energy per cycle. - **χ″(ω)** — the *absorptive*, *dissipative*, out-of-phase part: absorption coefficient, loss tangent, damping, energy dissipated per cycle. K–K says: **if you know the absorption spectrum at every frequency, the dispersion spectrum is determined — and vice versa.** You cannot engineer a material that absorbs strongly only in a narrow band without also producing a specific, calculable anomalous dispersion nearby. Every lossy medium is automatically a dispersive medium, in a quantitatively predictable way. This is not a small engineering convenience. It is a theorem about what causal linear systems can and cannot do. ## Why it is true — one sentence A linear response is causal iff its frequency-domain transfer function χ(ω), viewed as a function of complex ω, is **analytic in the upper half-plane** and decays at infinity. Cauchy's integral formula, applied to that analytic extension on a contour closed in the upper half-plane, with a small semicircular indentation around the real pole at ω′ = ω, produces exactly the two K–K integrals. Causality in the time domain becomes analyticity in the frequency domain, and analyticity pushes real and imaginary parts into a fixed integral partnership. Spelling out that one sentence rigorously is the purpose of this entire series. ## A worked teaser — the Lorentz oscillator A single damped Lorentz resonance gives the classical dielectric susceptibility ``` χ(ω) = ω_p² / (ω₀² − ω² − iγω) ``` Splitting into real and imaginary parts, ``` χ′(ω) = ω_p² (ω₀² − ω²) / [(ω₀² − ω²)² + γ²ω²] χ″(ω) = ω_p² γω / [(ω₀² − ω²)² + γ²ω²] ``` χ″(ω) is a peak centred at ω = ω₀ with width ≈ γ. χ′(ω) is the textbook anomalous-dispersion curve: it rises below ω₀, swings through zero at ω₀, goes negative just above, and returns to zero at high ω. Substituting χ″ into the K–K integral reproduces χ′ exactly — the two curves are locked together. Section 13 of the series does this integral explicitly. ## Sign and Fourier conventions K–K is almost always stated with the **physics** Fourier convention `e^{−iωt}`: a signal of positive frequency oscillates as `e^{−iωt}` and causality places the analytic extension of χ(ω) in the **upper** half plane (Im ω > 0). In engineering and signal-processing texts the convention is `e^{+jωt}`, which flips the sign in every exponential and places the analyticity region in the **lower** half of the `s = jω` plane. The two conventions give the same physics; only the sign of certain cross terms changes. This series uses the physics convention throughout, and flags the engineering counterpart where it matters. ## Roadmap of the series The series is divided into three arcs. **Foundations of complex analysis (Chapters 2–9).** The reader is assumed comfortable with complex numbers and basic calculus but not with residue calculus or Hilbert transforms. Chapter 2 places K–K in historical context; Chapter 3 introduces holomorphy; Chapter 4 covers the Cauchy–Riemann equations and the link between harmonic conjugates and K–K; Chapter 5 builds up contour integration; Chapter 6 states the Cauchy theorems that are the backbone of the derivation; Chapter 7 covers residues; Chapter 8 introduces the Cauchy principal value and the Sokhotski–Plemelj identity; Chapter 9 proves Jordan's lemma, the "close the contour at infinity" move. **The K–K theorem itself (Chapters 10–13).** Linear response and causality; analyticity in the upper half-plane; derivation of the dispersion relations; subtractions; sum rules (f-sum rule, superconvergence). **Applications (Chapters 14+).** Optics and refractive index; dielectric spectroscopy; network theory and S-parameters; acoustics and mechanical compliance; magnetic susceptibility; high-energy scattering amplitudes; practical reconstruction from band-limited data. ## See also - [02_historical_context.md](02_historical_context.md) - [03_complex_functions_analyticity.md](03_complex_functions_analyticity.md)