Stress Cracking in Polycarbonate: The Machining Causes Nobody Warns You About

Polycarbonate (PC) is an amorphous engineering thermoplastic with a glass transition temperature of approximately 147 °C, tensile modulus near 2.3 GPa, and tensile strength in the range of 60 MPa to 70 MPa for unfilled injection-moulded grades. The notched Izod impact strength of unreinforced PC under ASTM D256-23 is often reported above 600 J/m, but this value is thickness-dependent and drops sharply when a machined surface replaces a moulded skin. In medical device housings, fluid handling manifolds, optical inspection covers, and transparent machine guards, PC is frequently machined from extruded sheet, cast rod, or molded blanks. That secondary processing introduces failure modes absent from moulded-only parts. The phrase “stress cracking in polycarbonate” is widely associated with chemical exposure, but the machining causes are less commonly documented. Stress cracking after machining arises from the interaction of residual tensile stress, geometric stress concentration, thermal history, and contact with process fluids. A part can crack days after machining without externally applied load because the stored energy in the machined surface layer is sufficient to propagate crazes once a critical flaw depth is reached.

Standards for raw material characterisation do not directly measure machined-surface integrity. Table 1 lists relevant test methods used when evaluating machined PC. Values obtained from these tests are geometry- and preparation-dependent; machined specimens must be prepared according to ISO 2818:2018 and compared with annealed control pieces. When the machined edge differs from the annealed edge in tensile elongation at break, the process is considered suspect in many aerospace and medical device procurement specifications.

How Does Tool Contact Transfer Residual Tensile Stress into a 0.5 mm Deep Surface Layer?

Cutting polycarbonate is not a clean shear process. The tool edge compresses the polymer ahead of the rake face, generates friction at the flank face, and leaves a plastically deformed surface layer that does not fully recover. In unfilled PC, the affected depth is commonly 0.2 mm to 0.5 mm depending on tool sharpness, feed, and cooling. Within this layer, polymer chains align along the cutting direction and then solidify under constraint as the local temperature falls from the cutting-zone peak toward ambient. The outer surface ends in a state of tensile stress, while deeper material balances this with compressive stress. When the part geometry includes a drilled hole, an O-ring groove, or a sharp internal corner, the stress field redistributes. A small root radius produces local stress intensification that can exceed the craze initiation stress of PC even without chemical contact. Machinists often dress a tool to keep it cutting, but a worn flank land increases friction and raises surface temperature. A dull tool can raise local temperature above the glass transition and then quench the polymer, producing a surface skin with higher chain alignment and greater residual stress than a sharp tool would create. The effect is not detected by dimensional inspection because the damage is sub-surface.

Spindle speed, feed per tooth, and edge preparation control this damage layer. Process windows for unfilled and glass-filled PC are grouped in Table 2. The greatest error on production CNC machining centres is excessive dwell at hole entry and exit. A three-axis vertical machining centre with a 24,000 min^-1 spindle may be run at conservative cutting speed but still generate cracking because the program retracts the tool without maintaining chip load; this causes a momentary burnishing action. Tool suppliers specify positive rake angles of 5° to 15° for solid-carbide end mills in PC. The clearance angle at 10° to 20° prevents flank rubbing. Surface speeds for carbide tooling are normally set between 150 m/min and 300 m/min for unfilled extruded PC; glass-filled grades require the lower end. Feed per tooth values below 0.05 mm generate excessive frictional heat, while values above 0.15 mm produce mechanical tearing. The usable window in thin-wall sections below 3 mm narrows because less material is available to carry heat away, and thermal expansion during the cut can create immediate residual stress gradients. The recommended approach is to adjust radial engagement instead of reducing spindle speed alone. Constant chip thickness, climb milling, and high-pressure air or inert mist cooling are more stable than flood coolant for PC. Table 2 is a starting point only; validation must be performed on the specific machine because spindle runout, tool holder balance, and fixture rigidity alter the effective limits.

Annealing at 121 °C Does Not Uniformly Erase Milling-Induced Stress

Annealing is the standard countermeasure, but its limits must be understood. Below the 147 °C glass transition, PC chains have limited mobility. An air oven soak at 121 °C for 2 h to 4 h allows relaxation of the highest-stress surface chains but does not necessarily anneal the core. The rate of thermal diffusion and creep relaxation are thickness-dependent; a 25 mm thick blank may require several hours just to reach a uniform core temperature. If the part is heated too quickly, the surface expands faster than the core and introduces new transient stress. Heating and cooling ramps of 5 °C/min are commonly specified for PC stress relief. An oven profile deviation of ±5 °C around 121 °C is considered the maximum allowable in stress-relief cycles for optical PC parts; lower temperatures reduce relaxation, while higher temperatures risk dimensional distortion and optical haze. The part should be supported on a flat, non-abrasive fixture and not stacked. Parts with internal threads, undercuts, or thick-to-thin transitions should be annealed before final finishing, because post-anneal machining re-introduces stress at the newly cut surfaces. Polariscopic inspection after annealing often shows that birefringence at a sharp corner is reduced but not eliminated. Complete thermal stress relief in PC would require temperatures approaching the glass transition, which risks dimensional instability and optical distortion. Therefore, annealing is a partial mitigation, not a cure. Validated cycle parameters should be recorded as an oven profile with thermocouple placement in the largest section, not from the air temperature display alone.

Coolant selection interacts directly with residual stress. Polycarbonate is not chemically resistant to all water-miscible machining fluids. The problem is not only the base fluid; additives such as amines, glycol ethers, borate esters, and aromatic sulfonates can plasticise the surface and lower craze initiation stress. In a machined part with high residual tensile stress, a cutting fluid that is harmless to an unstressed moulded part can produce environmental stress cracking on a stressed surface. The appropriate screening method is to expose a bent strip to the candidate fluid under constant strain according to ISO 22088-3:2002 or ASTM D543-21. The strain level should match the worst-case residual strain at the machined feature, typically between 0.5% and 1.0% for critical features unless field measurements indicate higher local strain. A blank test without fluid but with the same surface preparation should be run in parallel to separate mechanical residual stress from chemical stress cracking. If the fluid causes crazing within 48 h, it should not be used on the machine. The same logic applies to cleaning agents, rust inhibitors, tapping compounds, and way oils. Isopropyl alcohol is frequently implicated because it is used as a cleaner for PC but is a mild stress cracking agent; a heavily stressed machined feature may craze within minutes of wiping with isopropanol. Even fast-drying contact cleaners should be evaluated before use on machined PC. Production lines that machine PC dry often fail later because the part is placed in a foam-lined tray still wet with a rust inhibitor transferred from the operator’s gloves.

When Glass-Reinforced Grades Convert Machining Stress into Fast Fracture Instead of Ductile Deformation

Glass-fibre-reinforced PC grades, typically 10% to 30% by weight, exhibit higher tensile modulus and lower thermal expansion than unfilled PC. They also have lower notched impact strength and greater susceptibility to interface-driven microcracking. During machining, the cutting tool fractures fibres at the surface rather than cutting them cleanly. The exposed fibre ends and surrounding matrix create discontinuities that concentrate stress at the microscopic scale. Differential contraction between glass fibres and the PC matrix after cooling adds a second residual stress field. A process that produces transparent, crack-free edges in unfilled PC may generate visible fibre pull-out and white microcrazing in 20% glass-filled grade. The main adjustment is to lower feed and increase tool sharpness; solid-carbide tools with positive rake remain appropriate, but surface speed should be shifted toward the lower end of the unfilled range. Annealing is still required, but it cannot restore the lost impact energy at the fibre–matrix interface. Comparative notched Charpy tests under ISO 179-1:2023 or Izod tests under ASTM D256-23 on identically machined specimens are used to quantify embrittlement. The fracture appearance changes from ductile tearing with stress whitening to smoother, more brittle fracture surfaces at lower impact energies. Optical microscopy at 50× to 200× magnification is required to distinguish between surface craze networks and fibre-matrix pull-out, because both appear as white marks to the unaided eye.

Non-destructive evaluation of machined PC is limited to transparent grades. Polariscopic inspection between crossed polarizers reveals birefringence; higher fringe density corresponds to higher residual stress difference. It is semi-quantitative and cannot be used on opaque glass-filled parts unless a photoelastic coating is applied. For opaque grades, ultrasonic or X-ray computed tomography can detect gross voids and delaminations, but published data for detection of fine microcrazes in PC at production speeds is limited. Destructive examination of cross-sections must avoid introducing new cracks. A diamond saw with low feed and water-free cutting is followed by polishing to 1 µm. Etching is generally not required because PC is amorphous; craze fibrils are visible under transmitted or reflected light if the section is polished without smearing. The presence of a craze network extending deeper than 0.5 mm from the machined edge after annealing is a negative indicator for long-term load-bearing service. Process validation should include witness coupons machined with the production tool and tested for tensile elongation at break under ISO 527-2:2012 or ASTM D638-14. The location of fracture indicates the dominant stress concentration: failure initiating at the machined edge rather than in the gauge section confirms that the surface condition controls strength.

Cryogenic Deburring, Edge Radius, and Dust-Free Fixtures in Production of Optical PC

Finishing operations after machining are a separate source of stress cracking. Manual scraping with a steel blade creates parallel microgrooves that behave as crack nuclei. Cryogenic deburring of machined PC can remove friable flash but creates a risk of low-temperature embrittlement if the part temperature falls below the ductile-to-brittle transition. In practice, tumbling PC in cryogenic systems is rarely used for precision optical components because surface microcracking can occur. A controlled edge break of 0.1 mm to 0.2 mm using a rotary abrasive brush with fine aluminium oxide or silicon carbide media is more reproducible. The brush speed must be low enough to avoid generating frictional heat. Fixtures with knurled aluminium jaws or hardened steel set screws create point loading that can craze the surface. Polyurethane or PTFE contact pads are used to hold PC parts without inducing additional residual stress. Vacuum fixtures are preferred for flat panels, but the vacuum pressure should not distort the part locally. Clamping force on injection-moulded blanks must be lower than that used for metal workpieces; a rough check is to clamp only enough to prevent movement and then reduce by 20%. The operation should be performed in a dust-controlled environment because airborne particles trapped under the fixture can create micro-indentations. In optical applications, any surface contact mark is a potential craze initiation site.

Design reviews of machined PC parts should address stress risers before machining begins. Holes should be drilled with a two-step process when the diameter exceeds 6 mm: a smaller pilot hole followed by reaming to final size reduces the residual stress around the bore. Threaded features should use coarse threads and oversized inserts rather than thread-cutting taps when repeated assembly loads are expected. Thread-cutting taps create sharp roots and local stress concentrations; installing a metal threaded insert distributes the assembly load over a larger area. Press-fit pins into PC bores are a frequent source of delayed stress cracking because the hoop stress from interference fit adds to the machining stress. Interference levels above 0.5% of the hole diameter are considered high risk for unfilled machined PC; this value should be confirmed by creep data and service temperature. Where a press fit is unavoidable, the hole should be machined with a final reaming operation, annealed, and then tested under the maximum service temperature and chemical environment. Sharp internal corners should be designed with a radius of at least 0.5 mm and machined with a ball end mill that maintains the radius through the full depth. A drawing that specifies a 0.5 mm radius but programs a roughing cutter that leaves a 0.2 mm radius at the root creates a stress cracking risk that cannot be corrected by a simple finishing pass.