Introduction: The “color” of a pigment is not as simple as what meets the eye
The “ultimate goal” of paint color matching: Given a target color (e.g., the customer’s color chip) — the computer “first-principles” prediction of which pigments are needed and in what amounts — the mixed color under any light source (D65/A/TL84) and at any angle (15°/45°/110°) matches the target color (ΔE<1.0). This seemingly simple problem is extremely complex mathematically because a pigment’s color is not a “scalar” (a single number) — but rather a continuous spectrum from 380–780 nm (R(λ)). Under different light sources (D65 vs A) — the same R(λ) integrates to different tristimulus values (XYZ) — this is the mathematical root of metamerism. The Kubelka-Munk theory (1931) quantified a pigment’s R(λ) into two optical constants (K absorption / S scattering), laying the theoretical foundation for computer color matching in paints.
I. Kubelka-Munk Two-Constant Theory
The core equation of K-M theory K/S=(1-R∞)²/(2R∞)R∞=reflectance of an infinitely thick coating——K=absorption coefficient (mm⁻¹)——S=scattering coefficient (mm⁻¹). Pigment mixing——(K/S)mix of the mixture=Σci×(K/S)i (linear superposition)——ci=concentration of pigment i——this is the “linear assumption” of pigment mixing In practical applications, it holds basically true (error <5%) making K-M theory the mathematical foundation of computer color matching for coatings.
The “two constants” K and S of the K-M theory are independent of each other: titanium dioxide (white/high S/strong scattering) + carbon black (black/high K/strong absorption) — the (K/S)mix after mixing the two — K comes from carbon black — S comes from titanium white — the mixed color is the “optical average” of white and black. But the limitations of the K-M theory: (1) it assumes pigment particles are isotropically and uniformly distributed in the coating (actually sprayed pigments have orientation/distribution gradients); (2) it ignores the porosity/(PVC effect) of pigments (pores are an additional scattering source/not considered by the K-M model); (3) it completely fails for effect pigments such as metallic/pearlescent pigments (the color of effect pigments is specular reflection + interference/non-absorbing scattering) — requires multi-angle spectroscopy + radiative transfer model to replace K-M.
FAQ
Q1: Why is the error of K-M theory particularly large for carbon black color matching?Carbon black has an extremely high absorption coefficient K (>1000 mm⁻¹, far higher than other pigments) — a tiny weighing error (±0.01%) can cause a significant change in (K/S) (>10%) — the lightness/darkness of the gray seriously deviates from the target. The “accuracy” of carbon black color matching is the most difficult in coating color matching — it requires high-precision automatic dosing (accuracy ±0.1g) + “robustness analysis” of the color matching software.
Q2: How is the “Flop Index” of the multi-angle spectrophotometer (MA-T12/12 angles) calculated?Flop Index=2.69×(L15°-L110°)^1.11/(L45°)^0.86——Measures the “color flop” intensity of metallic flake coatingsThe higher the Flop, the more pronounced the silver effect / the stronger the metallic feel. The Flop Index is a standard color quality control parameter for automotive metallic paintsThe Flop value of metallic flake coatings must be controlled within ±2 of the standard panel.
Q3: Why is AI color matching (CNN/deep learning) more accurate than traditional K-M? The K-M model is a linear model that assumes the K/S of pigment mixing can be linearly superimposed—but in reality, the distribution, orientation, and scattering behavior of pigments in the coating are nonlinear. CNN deep learning—”learns” the nonlinear laws of pigment mixing from tens of thousands of actual color data of historical formulas—directly predicts formula → color—without assuming K-M’s linearity—prediction accuracy (ΔE1.0). But AI requires large amounts of high-quality training data; in “small color spaces” lacking training data (such as special metallic flake effects)—AI is inferior to K-M.
Q4: The “Formula Cost Optimization” function of color matching software?The same target color—there may be multiple pigment combinations that can achieve it (e.g., red—can use iron oxide red (low cost) or DPP red (high cost) + organic pigment (vivid))—among all formulas that are “color qualified (ΔE<1.0)", the color matching software automatically selects the formula with the lowest total pigment cost and sorts them from low to high cost—for the formulator engineer to choose. The cost difference—low-cost pigment combinations may be >80% cheaper than high-price combinations—in mass production—formula cost optimization is the core source of enterprise profit.
Q5: The “Five-Dimension” Color Difference Management of Metallic Paint Color Matching — Why Is It More Than 10 Times More Complex Than Solid Color Paint? The color of metallic paint changes with the observation angle — (1) 15° (specular reflection angle) — the aluminum flakes’ reflection is “bright/white”; (2) 45° (front view) — the main hue of the color; (3) 75°-110° (side view) — dark area — the “depth” of the color. In addition, the (4) particle size of the aluminum flakes and (5) directional alignment (Flop) are also required — simultaneous matching of 5 dimensions — is the most challenging part of color matching. Multi-angle spectrophotometers (5-12 angles) are essential equipment for metallic paint color matching.
Q6: The role of “Formula Robustness Analysis” (Monte Carlo simulation) in color matching?Weighing error——Pigment A target amount 10.0g——actual weighing may be 9.9-10.1g (±1%)——this tiny error in a “sensitive formula” (e.g., containing carbon black 0.05g/trace amount/weighing error may be >100%)——can cause color shift ΔE>2. Monte Carlo simulation——randomly generates >1000 sets of “combinations of weighing errors” calculates color ΔE under each error combinationstatistically analyzes the distribution of color shiftidentifies “sensitive pigments” engineers can adjust the formula (e.g., replace sensitive pigments with more stable ones)——reduce color sensitivity to errors.
Q7: Why does the “Light Booth” in the color matching center have multiple light sources?The light booth contains three light sources: D65 (daylight/6500K), A (incandescent/2856K), and TL84 (fluorescent/4000K)—(1) Rapid screening for metamerism—a color matches under D65—but “suddenly differs” under A light, which is metamerism (detected at the color matching stage—rather than the customer complaint stage); (2) Different countries have different standard light sources—China standard D65, Europe D65, US D65+A—multi-light-source evaluation is needed to match colors for global customers.
Q8: How does the “batch management” of coating colors ensure consistency between the 100th batch and the 1st batch?Each batch of coating retains the “standard color panel” from the first production batch as a physical color panel (permanent archive/light-proof/constant temperature and humidity). Each batch is compared with the standard color panel using a spectrophotometer — ΔE<1.0 = qualified. The spectrophotometer requires daily calibration (standard white plate + standard black plate) — temperature/humidity drift → the photometer “drifts”, color detection data shifts overall by >ΔE 0.2 — inter-batch color consistency will be “falsely high” or “falsely low”.
Q9: Global sharing of a cloud-based formula database for digital color management?Paint companies (e.g., PPG/Axalta)——color matching centers around the world share one “cloud-based formula database”The colorist in Paris enters a new formula in the morning——the colorist in Shanghai can call it up in the afternoon——globally synchronized. The cloud-based “data security” formula database is the company’s “core intellectual property” and must have high-strength encryption + access control + log auditing to prevent formula leakage to competitors.
Q10: Future “Unmanned Color Matching Laboratory” — Will robots + AI replace colorists?Robots automatically weigh pigments + automatically mix + automatically draw down → spectrophotometer automatically tests → AI judges whether the color is qualified → if not qualified → AI automatically calculates the correction plan → robot executes the correction → re-test → qualified → AI outputs the formula. The entire process requires no human intervention and runs 24h/day — color matching efficiency is >10 times that of manual work. The investment in an unmanned laboratory (>2 million RMB) is suitable for large coating enterprises and chain color matching centers. For small and medium coating factories — currently still mainly “manual + semi-automatic”.
Related Reading
Summary
The scientific foundation of coating color matching—the Kubelka-Munk two-constant theory (K absorption / S scattering)—quantifies pigment reflectance into linearly additive optical constants. Multi-angle spectrophotometers (5–12 angles) are essential equipment for metallic paint color matching. AI deep learning is gradually replacing traditional K-M—under “large color space / big data” scenarios—significantly improving color matching accuracy (ΔE < 0.5). Kexin New Materials provides customers with comprehensive color matching services and digital color management technical support.