How to Fix Flow Marks: Mold-Side Causes and Surface Fixes

A glossy ABS trim panel for an automotive interior program came off the press with a dull ripple band 0.4 mm deep across the last 100 mm of fill. The first containment decision was a heat-gun pass on 1,800 parts, and 11 percent of them warped out of tolerance and died anyway. The customer inspection rejected every shipment after that for two weeks until the root cause was closed with a 15 °C mold temperature step and a widened gate land. Flow marks are not a beauty-shop problem. They are written into the fill history, and only a mold-side change removes them for good.

Flow marks are repeating waves of gloss and dullness aligned to the fill direction, with a period of 0.1 to 2 mm and a depth of 0.02 to 0.3 mm. The pattern is a record of the melt front stalling and restarting. Splay fans from the gate as irregular streaks and points to gas. Jetting lays a solid snake of frozen material on the surface. Weld lines mark one seam where two fronts met, not a rhythm of ridges, and that seam belongs to the weld line guide.

The mold defect troubleshooting hub sorts these five look-alikes in one table. This guide owns the flow-induced wave and hands you the full mold-side toolkit: the temperature threshold, the speed window, gate land rules, section transitions, polish direction, texture depth, and the order-matrix that tells you which lever to pull first.

The Snapshot

• Wave period of 0.1 to 2 mm. Depth above 0.02 mm is readable under a 60 degree gloss meter, and above 0.1 mm is visible under normal shop light.
• A 10 to 20 °C mold step is the first lever. Most waves collapse or disappear once the melt front stays 5 to 10 °C above the freezing band of the resin.
• Gate land at 50 percent of local wall, minimum. A land thinner than half the wall shears the skin and can jet the melt across an open face.
• Texture hides what you cannot remove. SPI SP2200 bead blast or a 12 µm etched pattern scatters light so waves under 0.3 mm stop reading on the part.

Not Splay, Not Jetting: The Flow-Mark Family

Buyers group every surface complaint under flow marks, so the vendor chases the wrong defect for a month. Separation happens in two minutes on the inspection bench. Tilt the part under a single lamp and look at the pattern. A regular rhythm of ridges running with the fill is a flow wave. A silver fan with random streaks is splay, and that mechanism belongs to the splay and silver streaks guide.

Three look-alikes need naming before the fix list starts. Jetting is a snake or pigtail frozen into the surface near the gate, caused by a free jet that solidifies before the front arrives; the jetting prevention guide covers that mechanism. A weld line is one dull seam at a feature where fronts collide, not a repeating band. Die lines and drag marks are scratch geometry from ejection and do not follow fill direction at all.

The flow mark family has three accepted sub-types, and all three are the same stall mechanism with different spacing. The classic wave is a periodic gloss ripple, usually radial from a near-side gate or banded across a long fill path. The stagnation wave appears where the front slows against a blind corner or an isolated core. Cold-slug tails are the dried first-gram of the shot carried into the cavity as a tail or swirl with a dead, matte surface.

Why does the same defect read differently on two parts? Visibility is an optical problem, not a geometry problem. A wave with a 0.05 mm depth scatters light on a glossy surface and disappears on a textured one. That is why half of the fixes later in this guide control the light instead of the steel. But the honest sequence is to reduce the geometry first and mask the residue second, because a masked wave still carries a 2 to 5 percent reject exposure at cosmetic inspection.

Fountain Flow and the Frozen Layer

Melt advances through the cavity like a fountain. Material at the center of the channel stretches forward to the surface at the front, then stops at the wall. Against steel held 80 to 140 °C below the melt, that arriving material freezes in a fraction of a second. Within 0.1 to 0.5 seconds of wall contact a skin exists, and from that moment the cavity sees three zones: solid skin, a shearing transition layer, and a molten core.

Waves are a frozen-layer problem. When the front loses heat faster than the pressure delivers material, it stalls. The stall lets the skin thicken, and the next increment of melt buckles sideways against that wall instead of advancing. The restart buries the buckle and the ridge. A stall and restart cycle, repeated every few milliseconds, prints the periodic ripple into the skin you inspect later.

Two variables set whether the front recovers or keeps stalling. The first is the temperature of the melt front itself, not the barrel reading. The second is whether the wall section keeps the core molten long enough to bridge the next corrugation. A 1.5 mm wall holds a molten core far longer than a 0.8 mm wall, which is why thin-wall parts wave first at the last-fill corner and thick parts wave near the gate where orientation relaxes.

Thermal conductivity of the steel decides how violently the skin forms. Tool steel moves heat at 25 to 45 W/m·K. A blocked waterline, a region boundary on an unheated mold face, or a bare steel insert next to a hot one creates a local cold patch that reads 5 to 15 °C below the setpoint. The wave locks to that patch on every shot. Aluminum bases at 120 to 160 W/m·K cool almost evenly and rarely wave at all, which is why the defect is largely a steel-tool problem.

The mechanical consequence is worth repeating in buyer language: the pattern repeats across the whole part because it repeats every cycle. It is not a one-shot event, it cannot be polished off the part, and it cannot be reworked. Every wave part is either accepted or scrapped, and the decision cost is a scrap rate, not a labor number.

Temperature Thresholds: Melt Front First

The first documented fix for a classic wave is to raise the mold by 10 to 20 °C. The number is not folklore. It lifts the melt-front temperature by roughly the same amount, because the front loses less heat per step of advance when the steel is closer to the freezing band. In practice, most parts clear the visible band between the first and second 10 °C step: PP moved from 20 to 40 °C, ABS from 40 to 70 °C, and a semicrystalline PA6-GF30 from 80 to 100 °C.

Each resin has a practical ceiling above the band. PP stops rising around 60 to 80 °C because cycle time and warpage take over. ABS tolerates 60 to 80 °C. PC sits at 80 to 120 °C on the mold, POM at 80 to 120 °C, and PBT-GF at 60 to 120 °C. Above those windows the part sticks, sinks or distorts, and the wave trade stops paying for itself.

Melt temperature matters less than the front temperature, and chasing it by raising the barrel is the wrong knob. A hotter melt lowers viscosity a little and risks degradation gas above the resin window, turning a cosmetic defect into a chemical one. Step the mold first in 5 to 10 °C increments, log the wave depth against each step, then move melt only in 5 to 10 °C steps and only when the mold is already at its ceiling.

Watch the cycle math before approving a heat-only fix. On a 28 second cycle, adding 4 seconds of cooling time to hold a higher mold temperature costs 14 percent of the press capacity. On a program running 600,000 parts a year, that is 85,000 parts of lost output or one added press. The temperature question in front of a flow-mark defect is really a business question, and the answer changes if the wave band is tied to one corner: then the tool needs spot heating or gate work, not a hotter whole mold.

The Injection Speed Window

Speed is the second lever and it cuts both ways. A fast front generates its own heat: shear rates above 5,000 s-1 can lift the melt core 5 to 20 °C inside the cavity. A slow front spends its heat on the wall instead and arrives colder at the far end.

For the classic wave near the gate, slow down. A front that surges through an open section smears and buckles its skin, and each buckle becomes a ridge. Cut the first fill stage by 10 to 30 percent and re-shoot. Many ABS and PP cosmetic parts clear a visible band with one speed step, no temperature change and no steel.

For the stagnation wave near a blind area, speed up. The front dies because it runs out of distance before it runs out of heat. A faster second stage carries molten material past the pinch before the skin locks. Typical fix values: raise the mid-fill stage from 60 to 90 percent of maximum screw speed, or shorten total fill from 6 seconds to 3.5.

The profile template that clears most waves is three-stage. Enter the gate slowly at 10 to 20 percent of max speed, accelerate to 70 to 90 percent through the open body, then decelerate to 30 to 50 percent just before the wave zone. Adjust each step by 10 percent increments and photograph the same area at the same light angle after every change.

Do not chase the defect with speed alone if the wave follows the same feature on every shot. A feature-locked wave is geometry or temperature, not dynamics. Speed smooths a bad window; it never removes one.

Gate Geometry and Shear at the Skin

The gate is where the part surface is written. Melt squeezed through a thin land shears at rates past 50,000 s-1, the skin smears, and the part leaves with a swirl or a jet-like ripple fanning from the gate. A land shorter than 0.5 mm, or a cross-section below half the local wall, is the usual suspect.

The gate cross-section rule: keep the smallest gate dimension at 50 percent of the local wall or above, and keep the land length at 0.5 to 1.0 times the wall. A 2 mm wall wants a sub gate of 1.0 to 1.5 mm and a land of 1.0 to 2.0 mm. A flash-cut edge gate at 0.3 mm in a 1.8 mm wall will mark almost any glossy resin.

The land matters more than the opening. A gate opened to size with a short 0.3 mm land accelerates the melt into a free jet that folds on itself. Lengthen the land first, widen the opening second. A tab or fan gate that bleeds the melt sideways at 1 to 2 times the wall width kills gate swirl without a single process change.

← swipe to scroll →

Table note: dimension bands are typical mold-side values for general-purpose thermoplastics on cosmetic parts. Confirm against the resin datasheet and the gate selection rules in the injection mold gate type guide.

Gate position decides whether the wave ever becomes visible. A gate that fires straight across a 100 mm open face guarantees a swirl record. A gate that lands on a core, a rib or a thick boss buries the disturbance in turbulence the front can re-weld. Move the gate in simulation before you cut steel, and at trial move it to a witness tab before you scrap a cavity.

Wall Thickness Transitions

The front speed follows the section. When the path thickens, velocity drops and the skin thickens in the same instant. The next increment of melt arrives at a narrower throat than the frozen channel can carry, buckles, and prints a ridge. That is why so many wave bands sit exactly downstream of a section change.

The design rule every DFM review enforces: keep adjacent sections inside a 0.5 to 1.5 ratio. A 3 mm wall stepping straight to 1.2 mm fails the rule and waves at the pinch. Blend the change over a ramp of 4 to 6 times the thicker wall, which for a 3 mm section means a transition of 12 to 18 mm.

Ribs and bosses generate their own class of surface marks. Keep rib thickness at 50 to 60 percent of the parent wall and boss bases hollowed or cored, or the front will race the rib and skip across the face above it. The resulting band repeats on the glossy side and aligns perfectly with the rib trace, which is how you identify it at inspection.

When the part is already frozen in revision and the section cannot move, the tool can still help. Move the wave by repositioning the gate so the transition fills first and the thick zone packs behind it, or add a local mold temperature zone at the boundary. Both are cheaper than a part re-approval.

Polish Direction and Grain Flow

Polishing leaves a grain: a field of fine directional grooves. On a glossy face viewed under raking light, cross-grain diffusion mimics a flow wave and misleads the diagnosis. The shop rule is simple. Polish the cavity in the fill and ejection direction, never across it, and close the last abrasive pass toward the parting line.

The numbers that matter sit in the Ra band. Optical and piano-black faces want SPI A-1 to A-3, which is Ra 0.01 to 0.05 µm. At that level a 0.2 µm groove scatters light visibly. Hand-lap with the grain, drop one grade finer than the previous stone, and erase the previous grain fully before the next pass.

EDM adds a hidden second surface. The spark leaves a recast layer 1 to 5 µm deep with micro-craters 5 to 20 µm wide, and an unpolished EDM face replicates straight onto the part as a fine matte ripple. Sink EDM in the cavity always gets a grind-and-polish recovery pass after the burn, and the finish spec belongs on the drawing, not on a shop convention.

Direction matters most within 25 mm of the gate, where the skin is born and the front is slowest. A cross-grain polish right at the gate can hold a wave that no process change will clear, and the shop loses two weeks tuning temperature for a finish problem that takes four hours of re-polishing to erase.

Texture Masking: SP Standards and Blast

Sometimes the wave will not leave. A deep, slow-fill band tied to a functional section can survive every temperature and gate fix on the machine. At that point the correct engineering answer is to stop removing geometry and start controlling light. A textured surface scatters at thousands of pits, and a scatter site deeper than the wave hides the wave.

The practical threshold: a molded texture of SPI SP2200 or coarser bead blast, or an etched pattern 12 to 25 µm deep, masks flow waves up to roughly 0.3 mm with no visible read. The rule behind the number is simple arithmetic. Texture amplitude must exceed wave amplitude, or the wave still catches the light between the pits.

Texture has costs. Gloss falls from 85 to 95 units at 60 degrees on a polished face to 5 to 20 units on a blasted one, and the customer's colour, touch and mirror requirements all change with it. Etch depth beyond 40 µm starts to show its own directionality across long flow paths. Decide the texture on the drawing at program launch, never as a surprise after a failed cosmetic sign-off.

The specification discipline is covered in the mold texture spec guide: sample plates, the etch-vs-blast decision, and the depth numbers that survive 500,000 shots of glass-filled resin without polishing flat. A texture worn past 50 percent of depth lets the wave read through again at shot 400,000, which turns a masked defect into a tool refurbishment budget.

Troubleshooting Matrix: Signature to Cause

Read the pattern before you touch a setting. The wave's position, shape and spacing point to the cause, and each cause has exactly one first fix. Work the table top to bottom: process steps cost nothing, steel steps cost two to five weeks.

← swipe to scroll →

Table note: signatures assume a single-gate cosmetic face on semi-crystalline or amorphous general-purpose resin. Photograph the area before and after each change so the fix is attributable.

Vents and Witness Marks as Wave Makers

A vent that is too shallow makes the front hesitate exactly where the gas should leave. The stall prints a wave band along the vent line, and the shop calls it a fill problem. Vent depth belongs at 0.01 to 0.04 mm on the land depending on resin; the full depth-by-material table lives in the mold venting guide. If the wave follows the parting line for 50 mm or more, measure the vent with a feeler gauge before you touch the barrel.

Leader pin and ejector pin faces are small cold obstacles with big surface opinions. A pin standing 0.02 mm proud of the cavity face trips the skin and radiates a fine wave from the ring. Keep ejector and leader pin faces 0.01 to 0.02 mm below the cavity surface, and polish the pin head in the ejection direction. At trial, hunt proud pins with a blueing pass on the ejector side before blaming the process.

Gas expansion has a cosmetic twin here. Trapped air compressed ahead of a hesitating front superheats locally and leaves a dull halo, and the halo edge reads as a wave crest under the lamp. Venting problems and wave problems share the same last-fill geography, so clear the vent depth and land length first and re-judge the surface after 50 shots of stable heat soak.

When the Material Owns the Wave

Run the tool clean, still waving? Move the investigation to the hopper. Recycled content changes melt front behavior shot to shot. PCR streams vary 10 to 30 percent in melt flow index lot to lot, and a 20 percent swing in MFI is a 20 percent swing in wave amplitude at fixed settings. Lock PCR resins with a per-lot flow test before release.

Masterbatch and filler loadings break the front the same way. Carbon black above 3 to 4 percent, and calcium carbonate extenders above 10 percent, raise viscosity at the skin and promote hesitation bands on long fills. Glass at 15 to 35 percent brings its own rule: the fibers migrate out of the skin layer, so the wave amplitude grows with content and glass tracks appear beside the ridges.

Hygroscopic resins add a false-positive trap. PA6, PC and PBT at moisture above 0.02 to 0.05 percent throw fine streaks that sit on top of a genuine wave pattern. Run the dryer to a minus 30 to minus 40 C dew point and re-inspect before cutting a steel fix. The distinction matters because moisture streaks fan irregularly from the gate while stall waves stay periodic and aligned to the fill.

Prevention at DFM and Trial

Flow marks are cheapest to kill before the steel is cut. At DFM, hold sections inside the 0.5 to 1.5 ratio, keep the longest flow ratio under 150 to 200 to 1 for general resins, and put the gate on a feature, not on the optical face. Specify mold temperature zones where a fixed wave band is predictable: a bridged waterline over a blind corner costs 300 to 800 dollars in the tool and saves a re-polish cycle.

At trial, build the wave into the study instead of reacting to it. Run a mold temperature ladder at 10 C steps and a speed ladder at 10 percent steps, photograph the same grid of areas under fixed raking light, and record the depth reading with a stylus or optical profilometer at 0.02 mm resolution. The window you ship with the tool is the smallest combination that clears the band, not the one that just happens to work on the last five shots.

Acceptance numbers belong in the purchase order: wave depth under 0.02 mm on A-surfaces at the agreed gloss level, no visible band under a 1,000 lux lamp at 60 degrees viewing, and texture amplitude confirmed against the SP plate. A tool that meets all of them at T1 will meet them at 100,000 shots; one that passes on masked luck will not.

Flow Mark FAQ

Q1. Is a flow mark the same thing as splay?

No. Splay is gas streaking out of the melt and fans irregularly from the gate; its causes are moisture and shear gas. A flow mark is a periodic ridge aligned with the fill, caused by a stalling front. The tests differ: drying and venting cure splay, temperature and gate geometry cure waves.

Q2. How much mold temperature do I need to clear a wave?

Step 10 to 20 C from the current setpoint, in 5 to 10 C increments. Most amorphous cosmetic parts clear between 40 and 70 C mold temperature; semicrystalline parts need 80 to 100 C. Above the resin ceiling the trade turns into sink, warp and cycle loss.

Q3. Can a robot remove flow marks from the part?

No process removes the geometry after demold. Buffing melts the ridge and moves the wave 10 mm downstream. Painting at 40 to 60 µm film hides shallow waves under 0.05 mm and nothing deeper. Texture is the only reliable cosmetic recovery.

Q4. Why does only one cavity wave?

Because one cavity sees a different history. Check gate land variance of more than 0.05 mm between cavities, a pinched waterline, a proud ejector pin, or a runner tail that heats that drop. One-cavity waves are tooling or maintenance, never barrel settings.

Q5. Does faster injection always help?

It helps stagnation waves and hurts gate-entrance waves. Shear heating from a fast front raises melt temperature 5 to 20 C, which re-warms a dying front. The same speed smeared into a near-side gate prints the surge as ripples. Split the difference with a three-stage profile.

The Bottom Line

Flow marks are a frozen-layer story with four numbers that fix most of it: a melt front held 5 to 10 C above the freezing band by a 10 to 20 C mold step, a gate land at 0.5 to 1.0 times the wall with the opening at 50 percent of wall minimum, adjacent sections inside a 0.5 to 1.5 ratio, and a texture amplitude above 0.3 mm wherever a residue wave must be hidden. Read the signature first, work the matrix from process to steel, and confirm with a depth measurement at 0.02 mm resolution instead of a lamp and an opinion.

DieStrike builds and repairs cosmetic injection molds to IATF 16949 with ±0.005 mm mold accuracy, in-house texturing and polishing, 120+ machines and DFM feedback in 24 hours. Tool lead times run 2 to 5 weeks and standard components ship in 3 to 7 days. If a program is fighting a wave band on an A-surface, send the photos and the fill history and we will return the cause order and the fix list before the next trial.