Optical Plastic Parts: Controlling Birefringence

A molded lens, light guide or display cover can be geometrically perfect and still fail, because the polymer itself is optically anisotropic. Stretched chains and frozen-in stress make the refractive index depend on the direction of light travel, so a ray that enters as a clean wavefront leaves split into two components with a phase difference. That phase difference is birefringence, and it is the single most common reason a transparent part is rejected for optical use.

Figure 1. Optical Plastic Parts: Controlling Birefringence.

What birefringence actually is

Birefringence is reported as the difference between the refractive indices along two principal directions, Δn, or more practically as retardation R in nanometres: R = Δn × thickness. A retardation of 20 nm across a 2 mm lens is negligible for a display cover but unacceptable for a camera lens or a laser window.

Two mechanisms create it in a molded part.

• Flow-induced orientation. Polymer chains align with the melt front as the cavity fills. If the part freezes before the chains relax, the orientation is locked in.
• Photoelastic stress. Residual stress from packing, differential cooling or ejection also makes the material anisotropic. This is the same mechanism used deliberately in photoelastic stress analysis of transparent models.

The two add together, and in practice most molded-part retardation is a mixture. That matters because the fixes are different: orientation calls for slower filling and hotter molds, stress calls for better packing and more uniform cooling.

Measurement

You cannot control what you cannot measure, and birefringence is easy to measure once a bench exists.

• Polariscope with tint plate gives a fast qualitative picture: colour bands reveal the retardation distribution across the part, which is often enough to compare two gate positions.
• Compensator-based measurement gives retardation in nanometres at a point, using a Babinet or Senarmont compensator.
• Transmission ellipsometry or a dedicated retardance imager maps the whole part in one shot and is the practical tool for a production qualification.
• Photoelastic fringe count under crossed polarizers is the cheap shop-floor version: each full fringe corresponds to a known retardation for the wavelength used.

For most projects a map at three locations is enough: the optical zone, the gate area and the far edge. The gate area will always read high; what matters is the value inside the clear aperture.

Design levers

Optical quality is decided at the design stage, before any process window exists.

• Uniform wall section. A thick-to-thin transition creates a shear gradient and a cooling gradient at once. Both produce retardation. Where thickness must change, make the transition gradual and keep it away from the optical zone.
• Gate away from the optical path. Put the gate behind a rib, in a throw-away tab or in a masked border, and let the flow front arrive at the optical zone after it has settled.
• Avoid weld lines in the aperture. A weld line is a line of molecular orientation and often a visible optical step. Single-gate designs with a symmetric flow front are preferred, even at the cost of higher pressure.
• Flow direction versus polarization. For polarizers, wave plates and laser windows, the orientation direction relative to the polarization axis is part of the design. Specify it on the drawing so the molder can align the gate.
• Polish and surface quality. Mold surface roughness scatters light and confuses retardation readings. Optically polished inserts, not just SPI A-1, are needed for lenses.
• Insert design for cooling symmetry. A lens cavity cooled more on one side than the other will have a stress gradient regardless of process settings.

Process levers

Once the tool exists, the process window is about giving the chains time to relax before they freeze.

• Higher mold temperature. This is the most powerful lever. Running a PMMA lens at 90 °C rather than 60 °C mold temperature can halve retardation, because relaxation continues while the part is still soft.
• Slower injection. Lower shear rate means less orientation. The limit is the risk of short shots and hesitation marks.
• Profile the injection speed. Fast fill to the gate area, then decelerate as the front enters the optical zone. This is where a multi-stage velocity profile earns its keep.
• Lower packing pressure, longer hold. High packing pressure adds photoelastic stress. Pack to the minimum that holds dimensions, and extend hold time rather than pressure.
• Uniform cooling. Balance the circuit so both mold halves reach the same surface temperature. A 10 °C difference between halves shows up as a stress gradient through the thickness.
• Annealing after molding. A free-annealing step below the heat deflection temperature relaxes frozen-in stress and can cut retardation by 30 to 60 percent. It costs cycle time and must be validated for dimensional change, but for high-value optics it is often the cheapest fix available.

Drying and material handling

Optical parts punish moisture and contamination more than structural parts do.

• Moisture creates splay, silver streaks and micro-bubbles that scatter light. Dry PC to below 0.02 percent, PMMA and COC per the supplier data sheet.
• Contamination from regrind, dust or a previous colour shows up as visible specks. Optical work usually runs with zero regrind and a dedicated, closed material path.
• Residence time in the barrel must be short and consistent. A long dwell at high temperature yellows the resin and raises haze.

A worked comparison

Consider a 2 mm PMMA light guide with a 40 mm optical zone. The initial tool ran at 65 °C mold temperature with a 0.6 s fill and a single edge gate at the end of the light path. Retardation in the centre of the zone measured 120 nm and the part showed visible rainbow bands under a polarizer.

Three changes were made. The gate was moved to a masked end tab outside the viewing area, so the flow front entered the zone fully developed. Mold temperature was raised to 85 °C with two additional circuit passes near the optical face. Injection was profiled to fill at 35 mm/s up to 80 percent of the volume and then drop to 12 mm/s. Retardation in the same location fell to 38 nm, and the visible banding disappeared.

No change was made to the resin, the machine or the part geometry. The improvement came entirely from letting the melt arrive calmly and giving it time to relax.

Where birefringence is acceptable

Not every application needs a low-retardation part, and over-specifying is expensive.

• Display covers and windows: retardation below roughly 50 nm is usually invisible; the real risk is haze and surface waviness.
• Light guides: retardation causes polarization-dependent extraction, so uniformity matters more than the absolute value.
• Imaging lenses: 10 to 20 nm is a working target and may require annealing.
• Laser and sensor windows: often below 10 nm, and usually a specification on the drawing with a measurement method attached.

The specification should always name the measurement method, the location and the acceptance value. A birefringence requirement without a method is not a requirement.

Figure 2. Quick reference: the key checks and reference values from this guide.

Checklist for optical part design reviews

• Confirm the optical zone on the drawing and keep gates, ejector marks and weld lines outside it.
• Check wall uniformity through the optical path; taper any thickness change away from the aperture.
• Specify mold temperature range and the polishing requirement for optical surfaces.
• Decide whether annealing is part of the process route and allow for it in the cycle time and cost.
• Agree the birefringence specification with a method, a location and a value.
• Set a drying specification per material and lock the material path against contamination.
• Plan a first-article measurement on a polariscope before committing to production.

Birefringence is not a defect that can be inspected out at the end of the line. It is set by how the melt arrives, how fast it freezes and how the part is packed, which means the conversation has to happen at the tool design stage.