Color (Gardner) & Batch Consistency in Epoxy Resin: What Causes Yellowing

Liquid epoxy resin based on bisphenol A diglycidyl ether (DGEBA) is routinely specified with a maximum Gardner color of 1 for electrical laminates, clear coatings, and optical adhesives, while standard adhesive grades may permit Gardner 3. The primary test method is ASTM D1544-04(2018) using comparison tubes numbered 1 to 18. Batch consistency is particularly difficult because the visual comparator masks small drift below a full Gardner unit, yet downstream formulation and curing can detect that drift as haze or yellowing in clear films. The following sections examine raw material carryover, oxidative side reactions during synthesis, stripping damage, metallic residues, diluent blending, storage headspace, and measurement variance as the main contributors to Gardner color shift.

Raw Material Chromophore Load and Bisphenol A Purity

Before reactor charging, bisphenol A quality fixes a chromophore floor. Oxidized bisphenol A fractions and trisphenol impurities absorb in the 350–450 nm region and survive glycidyl etherification. Incoming bisphenol A is not typically judged on the Gardner scale because the value is below the resolution of the comparator; instead, a Pt-Co/APHA method such as ASTM D1209-05(2019) is used. A commercial QC specification may set melt color at or below 25 APHA, with release limits tied to the finished resin specification. When a railcar or supersack of bisphenol A drifts above that threshold, the resulting liquid resin commonly moves from Gardner <1 to Gardner 1–2 without any reactor upset. Low-isomer bisphenol A with limited ortho-substituted material reduces this sensitivity, but only if the solid flake or prill storage hopper is inerted and kept below 40°C to avoid autoxidation of the surface. Sourcing changes that lower bisphenol A purity below 99.5% can therefore create batch-to-batch Gardner fluctuations even when reactor conditions remain unchanged.

How Do Epichlorohydrin and Caustic Purity Enter the Color Equation?

Epichlorohydrin is not usually judged by visual color because it is a low-viscosity chlorinated organic, but it contains chloropropenes and other unsaturated by-products that can oligomerize under alkaline conditions. Finished epichlorohydrin with a purity below 99.8% supplies enough allyl chloride and dichlorohydrin to produce yellow resin after dehydrohalogenation. Sodium hydroxide quality also matters; diaphragm-grade caustic containing iron or chlorate above 10 mg/kg can darken the resin during the digestion step. These impurities are not visible in the clear caustic feed, so a batch can fail color even when bisphenol A and the main reactor profile are unchanged. Analytical tracking by gas chromatography for epichlorohydrin and by ICP-OES for iron in caustic provides a raw-material fingerprint that is correlated with final b* values.

Inside a 20 m³ glass-lined batch reactor, the etherification and dehydrohalogenation sequence is exothermic, and temperature excursions above 65°C accelerate oxidative coupling of phenolic intermediates into quinoid species. The reaction mass is held under a nitrogen pad, but transient air ingress during sodium hydroxide addition through a dip pipe introduces dissolved oxygen into a high-pH aqueous phase. At pH >13 and a free phenolic content above 2 wt%, the oxidation rate is sufficient to generate absorbance in the blue end of the visible spectrum. Closed-loop control of the sodium hydroxide dosing rate at 48–50 wt% concentration prevents localized hot spots, while jacket coolant supplied at 20–30°C maintains the bulk reaction temperature below the point where base-catalyzed epoxy hydrolysis becomes significant. Batches that exceed 70°C for more than 15 minutes during the final digestion step routinely show a final liquid resin Gardner shift of +1 to +2 units after stripping, even though the intermediate reactor sample remained acceptable.

Epichlorohydrin Recovery Exerts a Thermal Color Penalty

Excess epichlorohydrin is removed by vacuum distillation, often in wiped-film evaporators operating at 2–10 kPa and internal jacket temperatures of 80–120°C. The thermal history in this step is the most common source of new color in low-color DGEBA. At the upper end of that range, residual alkaline salts and epichlorohydrin can polymerize via ring-opening and form low-molecular-weight oligomers with conjugated unsaturation. The wiped-film rotor also creates a thin film with a short residence time, but localized skin temperatures at the heated wall can exceed the bulk set point by 5–15°C if the feed is not evenly distributed. Production lots stripped below 80°C and below 5 kPa generally retain their pre-stripping color; lots forced through a fouled evaporator at higher steam pressure show a Gardner increase of +2 or more and a higher viscosity from oligomer formation. The recovered epichlorohydrin stream containing hydrolyzed glycerin should be monitored by gas chromatography, because recycled epichlorohydrin with more than 0.5% total chlorine-containing impurities reintroduces chromophore precursors into the next batch.

Salt Fines, Iron Carryover, and Neutralization pH

After the resin is washed with water to remove sodium chloride, the interface layer carries soapy phenate residues and metal complexes. Iron from piping or from the bisphenol A supply forms intensely yellow iron-phenate complexes under alkaline conditions; even 0.5 mg/kg of iron can shift Gardner in a water-white resin. Neutralization with phosphoric acid to pH 5–7 before filtration precipitates iron phosphate and reduces residual alkalinity. However, over-neutralization below pH 4 can hydrolyze glycidyl groups and generate chlorohydrin by-products that contribute to color during later heating. Plate-and-frame filtration through 5 µm pads removes precipitated salt but can introduce color if the filter media contains leachable organic sizing. Filtration at 50–60°C with a differential pressure below 0.3 MPa prevents channeling and reduces the need for repeated passes. Batches that leave the filter with residual moisture above 0.1% often develop haze on storage, which is distinct from Gardner yellowing but is frequently confused with it during batch release.

Where the liquid resin is blended with C12–C14 aliphatic glycidyl ether reactive diluents, the final Gardner color is controlled by the lowest-purity component. These diluents frequently arrive with Gardner 1–3 because their manufacture involves acid-catalyzed etherification and distillation cuts that retain light unsaturated by-products. A formulator specifying a blend of DGEBA and 15 wt% reactive diluent cannot assume the blend will meet Gardner <1 unless the diluent itself is certified below that value by ASTM D1544-04(2018). Blending at 40–50°C under nitrogen with low-shear agitation below 200 RPM limits air entrainment; dissolved air is not harmless, because subsequent storage under a partially filled tote provides oxygen for slow chromophore development.

If Headspace Oxygen Is Not Displaced Below 0.5% by Volume

Oxidative storage drift is the least visible cause of batch inconsistency because the same production lot can exit the plant at Gardner 1 and reach the customer at Gardner 2–3 after three months in a warm warehouse. Liquid DGEBA is not typically formulated with phenolic antioxidants, so autoxidation of low-molecular-weight impurities proceeds slowly at ambient temperature and accelerates above 35°C. Bottles and totes should be nitrogen-blanketed to reduce headspace oxygen below 0.5% by volume, and repeated partial removal from a container should be avoided because each opening refreshes the headspace. A 1000 L intermediate bulk container stored at 35–40°C with a 10% headspace and no nitrogen pad shows measurable Gardner drift earlier than a sealed 200 L drum stored at 20–25°C, reflecting headspace-to-volume ratio and oxygen transport at the liquid surface. Published quantitative shelf-life data for specific resin grades are limited, but production-scale release records show that oxygen-exposed retain samples routinely fail visual color specifications before nitrogen-blanketed retain samples from the same batch.

Color measurement itself contributes to batch variation. The visual Gardner method under ASTM D1544-04(2018) is comparator-based and depends on operator daylight perception, tube cleanliness, and light source; north daylight and CIE Illuminant C are not identical. For tight specifications, facilities supplement the visual method with an instrumental yellowness index per ASTM E313-20 and CIELAB b* measured in a 10 mm transmittance cell. This is not a replacement for the Gardner specification, but it detects batch-to-batch shifts of 0.2–0.5 b* units that are invisible to the Gardner comparator. Calibration with certified reference oils or glass filters is necessary, and the sample temperature should be controlled at 25°C ± 1°C because the refractive index and turbidity of viscous resin vary with temperature.

What Cured Networks Reveal That Liquid Gardner Ratings Do Not

Liquid resin color is a necessary but insufficient predictor of cured yellowing. A DGEBA batch can meet Gardner <1 yet develop visible ambering after cure with cycloaliphatic amine hardeners under accelerated aging at 85°C and 85% relative humidity for 1000 hours. Cured-network yellowness depends on hardener chemistry, tertiary amine accelerators, UV exposure, and oxidation of the cured matrix, not solely on the liquid resin. However, liquid-phase chromophores from oxidized bisphenol A can persist through cure and amplify the initial cured color. For applications governed by optical appearance, the resin should be screened by casting a 3 mm plaque and measuring transmission haze and b* before and after exposure per ASTM G154 or ASTM D4459. If the liquid Gardner is tightly controlled but cured color drifts across batches, the root cause is typically an additive, accelerator, or hardener lot, not the DGEBA itself.

Establishing Process Capability for Gardner Color in Multi-Tote Campaigns

Batch release data for liquid DGEBA over a 12-month period can be evaluated as a Gardner frequency distribution, but the ordinal nature of the Gardner scale causes poor process capability resolution. A more useful approach is to collect b* or yellowness index data on every batch and trend that against the visual result. When process capability is calculated on b* with lower and upper specification limits derived from the Gardner limit, process shifts from a raw material lot or evaporator fouling become visible before a Gardner failure occurs. In multi-tote campaigns, sampling every tote rather than compositing from the batch tank reveals filling-line contamination and headspace effects. Production experience shows that a batch tank can be within specification while a single tote deviates by more than 1 Gardner unit because of residual sanitizer or a defective nitrogen pad on the filling lance.