Polyethylene terephthalate (PET) is a semicrystalline polyester whose melt-phase behaviour is governed by intrinsic viscosity, carboxyl end-group concentration, diethylene glycol content, acetaldehyde concentration, and CIE L*a*b* colour coordinates. Transfer from a pilot extruder running at 250 kg/h to a continuous production line running at 2,500 kg/h frequently produces a pattern in which the pilot run succeeds but full production drifts on PET IV and colour. The drift typically appears as a decline in intrinsic viscosity from a specification set point of 0.80 dL/g to measured values of 0.72–0.76 dL/g and a simultaneous positive shift in yellowness b* from 0.8–1.4 to 2.8–4.6, determined in accordance with ASTM D4603-18 and ASTM D6290-19 respectively. These changes arise from thermal, hydrolytic, and thermo-oxidative chain scission, compounded by residence time distribution broadening, oxygen ingress, and retained volatile by-products in larger melt vessels.

Pilot Run Succeeds but Full Production Drifts on PET IV and Colour

The scale-up discontinuity is seldom caused by a single variable. In a 38:1 L/D co-rotating twin-screw extruder, pilot operation at 500 kg/h commonly permits average melt residence times below 5 min and barrel set temperatures of 260–280 °C. The same screw profile at 2,800 kg/h can produce local melt temperatures above 290 °C in the metering zone because viscous dissipation per unit surface area increases while the surface-to-volume ratio available for barrel cooling decreases. Intrinsic viscosity retention under nitrogen-blanketed pilot conditions often remains above 96%; under full production with intermittent vacuum venting and atmospheric hopper feeding, retention can fall to 88–94% within one shift. The loss mechanism is pseudo-first-order random scission at ester linkages, with reported activation energies in the 160–180 kJ/mol range under oxygen-deficient conditions. The resulting carboxyl end-group increase is measurable by titration, and residual acetaldehyde generation can exceed 1.5–3.0 ppm when downstream preform injection follows without adequate degassing, as measured by ASTM F2013-10.

Why Does Intrinsic Viscosity Retention Drop When Throughput Is Scaled from Pilot to Continuous Melt Conversion?

Hydrolytic degradation becomes the dominant pathway when pellet moisture content rises above 0.005 wt% entering the screw. Pilot dryers with a desiccant bed capacity matched to 300 kg/h can maintain return air dew points below -40 °C; full-scale hopper dryers handling 3,000 kg/h may experience channeling, uneven bed depth, and return air dew point excursions to -20 °C when the regeneration cycle lags moisture load. Each 0.01 wt% residual water hydrolyzes ester bonds and produces one carboxyl end group and one hydroxyl end group per scission event; because the carboxyl group acid-catalyzes further hydrolysis, the reaction becomes autocatalytic. Infrared moisture analysers calibrated to ASTM D4019-22 and in-line viscometers installed after the melt pump are the primary safeguards. If the full-production line lacks a -50 °C dew point nitrogen purge over the feed throat, oxygen and atmospheric moisture are drawn into the plastication zone through the hopper and side feeders, reinforcing the shift in molecular weight distribution.

Solid-state polymerization capacity can partially recover lost IV, but it cannot erase upstream colour damage. Continuous solid-state reactors operating at 210–225 °C under nitrogen can raise IV from 0.74 dL/g to 0.82 dL/g over 12–16 h; however, the acetaldehyde and colour precursors generated during melt processing remain in the pellet until molding. In bottle-grade production, the subsequent injection molding stage then regenerates acetaldehyde through thermal decomposition of vinyl ester end groups, making the final preform acetaldehyde level higher than the pellet value. This offset explains why a resin lot may pass pellet IV and pellet colour release tests yet fail downstream preform specifications under ASTM D6290-19 and ASTM F2013-10.

Thermal, Hydrolytic, and Thermo-Oxidative Degradation Pathways Compete in Large-Vessel Melt Processing

Thermal degradation in PET proceeds through random chain scission without oxygen and generates vinyl ester and carboxyl end groups. Hydrolytic scission inserts water across the ester linkage and produces carboxyl and hydroxyl chain ends. Thermo-oxidative degradation proceeds through radical abstraction at aliphatic glycol units, hydroperoxide formation, and decomposition to conjugated carbonyl chromophores, which are the primary chemical basis of yellowing. The three mechanisms are not independent; oxygen ingress accelerates radical formation, and the resulting carboxyl groups increase hydrophilicity and promote further hydrolysis. Thermo-oxidative yellowing may appear at oxygen concentrations as low as 20 ppm in the melt, which is below the detection threshold of many plant gas analyzers but sufficient to shift b* by more than 1.0 unit over extended residence time.

Diethylene glycol formation is a parallel side reaction that consumes hydroxyl end groups and creates ether linkages. At elevated melt temperatures above 285 °C, diethylene glycol content can increase by 0.2–0.4 mol% relative to the resin cook, reducing thermal stability and lowering crystallinity in later solid-state processing. The colour shift is not solely a surface phenomenon; the chromophores are distributed through the pellet cross-section when oxygen is present in the melt before pelletization. Die-plate polymer degradation, vent vacuum loss, and degraded stagnant layers along the extruder barrel all contribute to pellet-to-pellet b* variability under full production conditions.

When Full Production Colour Drifts from b* 1.8 to b* 4.5 Without Changes in Cobalt Toner or Optical Brightener Dosing

Colour drift in PET full production can occur while the colourant dosing system remains unchanged. Antimony trioxide catalyst residues, commonly present at 180–250 ppm as Sb, can undergo reduction to elemental antimony in oxygen-starved or high-temperature zones, producing grey to blue-grey discolouration. Iron from dryer internals, transfer lines, or wear surfaces at concentrations as low as 3–5 ppm accelerates oxidative degradation and shifts the melt toward red-yellow. Phosphoric acid stabilizer at 20–50 ppm as P is used to suppress antimony reduction, but excessive phosphorous can form visible haze and reduce clarity. Titanium dioxide, when used as a reheat additive, introduces scattering centres that change the perceived CIE L* value and interact with yellow chromophores, complicating colorimeter-based correction.

Cobalt acetate is added as a blue toner to mask yellowing, but colour compensation does not address the underlying molecular weight drift. Dosing above 40 ppm as Co can overshoot into green-grey and reduce brightness. The appropriate release test is ASTM D6290-19 using CIE Illuminant D65 and a 10° observer; this method separates yellowness index from subjective visual grading. Production lines that rely only on visual pellet inspection often fail to detect b* drift from 1.8 to 3.5 until the value is outside the downstream preform specification. Near-infrared spectrophotometers mounted above the pellet classifier can track C-H overtone changes associated with ester loss, but published data for this specific closed-loop compensation configuration is limited.

Scaling solid-state polymerization from pilot batch reactors of 1,000 kg to continuous units of 40,000 kg/day alters heat transfer, nitrogen distribution, and pellet bed residence time. Pilot reactors can maintain uniform pellet temperature within ±3 °C, whereas full-scale cylindrical SSP vessels may show radial temperature differences of 5–10 °C during start-up. These differences affect both IV build rate and colour stability. If the melt-phase resin before SSP carries high carboxyl end-group concentration, the SSP reactor must consume more residence time to reach terminal IV, increasing the risk of thermal yellowing. A production line that compensates by raising SSP temperature above 225 °C may accelerate IV build but simultaneously generate additional acetaldehyde and colour bodies, creating a release paradox in which pellet IV conforms while bottle colour fails.

Residence Time Distribution, Oxygen Ingress, and Catalyst Residue Interactions

Residence time distribution is a stronger source of full-production drift than nominal temperature set-point error. In a pilot extruder, the dimensionless residence time distribution may approximate a narrow band with a variance below 0.4. In a large production extruder, dead zones behind flight clearances, static mixing elements, screen packs, and melt pump bypass lines create a long tail in the distribution. A fraction of the melt remains above 285 °C for 10–15 min, while the average residence time is only 6 min. This tail fraction degrades, generates carboxyl groups and chromophores, and then blends back into the main stream, pulling the average IV downward and the b* upward. Melt filtration removes only the largest gel particles; soluble degradation products remain.

Oxygen ingress sources include rotary valve leakage, open hopper flanges, vacuum vent seal water, and the vacuum pump discharge. In full production, nitrogen blanketing is sometimes disconnected to reduce operating cost, which permits ambient oxygen to enter the feed throat and degas vent. At 280 °C, the melt dissolves oxygen rapidly, and the thermo-oxidative radical pathway becomes measurable even when the gas phase oxygen concentration is below 0.5 vol%. The result is not only yellowing but also an increase in carboxyl end groups, which lowers hydrolytic stability and reduces the effectiveness of cobalt acetate masking.

Can Colour Compensation With Cobalt Acetate Mask the Underlying IV Drift?

Cobalt acetate compensation can shift b* back toward neutral, but it does not restore molecular weight or acetaldehyde performance. In bottle-grade PET, the preform acetaldehyde limit is frequently set below 4 ppm in the sidewall and below 8 ppm in the neck area, and the colour specification may allow b* no higher than 2.0 for water packaging. When cobalt acetate is raised above 30 ppm as Co to counteract severe yellowing, the melt may appear acceptable under D65 daylight but fail under indoor lighting or in the presence of titanium dioxide. The underlying IV loss also weakens melt strength, reducing preform injection stability and increasing gate crystallinity. Thus, colour masking without IV control creates a non-conforming product that passes only the most superficial visual checks.

For food-contact PET, the finished resin is assessed under FDA 21 CFR 177.1630 and Regulation (EU) No 10/2011, with migration testing for acetaldehyde, antimony, and degradation products. These compliance frameworks do not specify a production-scale colour compensation limit, but they assume that polymer degradation is controlled by good manufacturing practice. A full-production line that shows isolated pilot success followed by IV and colour drift must revalidate its drying, residence time, oxygen exclusion, and stabilizer strategy using ASTM D4603-18, ASTM D6290-19, and ASTM F2013-10 rather than relying on periodic pellet inspection. Thermocouple arrays at the die, continuous dew-point sensors, and offline carboxyl end-group titration provide the necessary boundary data for corrective action.