Flex PCB Material Selection: Polyimide, Copper, Coverlay

A flex circuit is not "soft FR-4"

The most common starting error in flex selection is treating a flex board as a rigid board with a different substrate: finish the layout the rigid way, then ask whether it can be built on film.

The materials are solving a different problem. On a rigid board, the material's job is to hold shape — Tg, thickness, post-lamination dimensional stability all exist so copper and vias stay put. On a flex circuit, the job is the opposite: maintain continuity while the shape keeps changing. Copper takes tension and compression at the bend, the dielectric must not delaminate under strain, and the surface layer has to move with the bend without cracking.

That difference has a practical consequence. Flex reliability is largely fixed the moment the material stack is chosen; it is not recovered by a later process step. On rigid boards, electrical performance is defended through controlled lamination, imaging, drilling and plating. On flex, bend life is written into which copper foil was used, whether there is adhesive in the stack, and how the conductors are covered.

The deployed flex product page states the baseline: 1–8 layers, PI (polyimide) base material, 2 mil minimum trace width, 0.5–2 oz copper, PI/PSA coverlay, and bend radius ≥ 6× board thickness.

That last item deserves its own sentence. It is a geometric constraint, not a material performance figure. Once mechanical engineering fixes the available bend radius, your total stack thickness is capped in return — and total thickness is the sum of base film, copper, any adhesive, and the coverlay. Mechanical's bend radius is the first input to material selection, not the last acceptance check.

This article walks the chain in build order: PI base and adhesive system → copper foil type → coverlay → stiffeners → use-case classification (dynamic flex vs form-and-stay). Each link is described in terms of what it locks in.

Link 1: PI base material and the adhesive question

Polyimide dominates flex base material for a reason beyond bendability: its temperature capability lets flex circuits go through standard reflow. That gets taken for granted in selection, and it is precisely PI's decisive advantage over other flexible films.

The deployed materials library lists the polyimide systems in use with typical values (the page states explicitly that values are typical and that complete data should be taken from the manufacturer datasheet):

All four are tagged "temperature cycling" on that page, which tells you what actually drives PI selection: holding structural and electrical integrity through repeated thermal excursions. If your board must bend *and* cycle thermally — automotive, medical instruments, industrial sensing — PI is the start of the chain rather than an option in it. Grade-level availability for a given project requires engineering review.

What affects bend behaviour more than the grade name is whether there is adhesive between base film and copper.

Adhesive-based construction: a bonding layer (commonly epoxy or acrylic types) sits between PI and copper. Mature, lower cost, broadly available. The cost is that this layer is the most dissimilar material in the stack — its thermal expansion behaviour differs from both PI and copper, making it a stress concentration site under thermal cycling and bending strain.

Adhesiveless construction: copper is bonded directly to the PI with no third material introduced. Total thickness drops (which directly relaxes the ≥ 6× thickness bend radius constraint), the layers behave more alike, and thermal and dimensional stability improve. The cost is higher material price and a narrower set of available constructions.

The decision rule is short: static forming only, low layer count, cost sensitive → adhesive-based is sufficient; repeated dynamic bending, thin stack, thermal cycling → start from adhesiveless. Available constructions and project fit require engineering review.

Link 2: rolled-annealed vs electrodeposited copper

If the material chain could only get one decision right, it should be this one. Copper foil type carries the largest weight in dynamic bend performance, and it is routinely left unspecified in Gerber packages that state only a copper weight.

The difference comes from how the foil is made, which sets the grain structure:

Electrodeposited (ED) copper is grown by electrodeposition onto a carrier. Grains are columnar, roughly perpendicular to the foil plane. Fine on rigid boards, where copper never has to deform repeatedly. In a bend, that grain boundary orientation is unfavourable relative to bending tension, and cracks propagate along boundaries more readily under repeated flexing.

Rolled-annealed (RA) copper is produced by repeatedly rolling and annealing, which flattens grains into a layered structure elongated in the plane of the foil and grows them during anneal. Aligned with the bending strain direction, the copper accommodates deformation through slip within grains rather than opening along boundaries.

The conclusion is industry consensus, not a vendor position: use RA copper in dynamic bend zones; ED copper is acceptable on flex that stays put after assembly. Note that this is independent of copper weight — the flex page range is 0.5–2 oz, but 1 oz ED and 1 oz RA do not behave the same in a bend.

Copper weight has its own effect, in a straightforward direction: thicker copper means higher surface strain at the bend. Bend zones should use the thinnest copper that the circuit's current requirement allows, with heavier copper confined to non-bending areas. When one board must carry substantial current *and* flex repeatedly, that is a genuine design conflict; the resolution (zoned copper weight, splitting conductor paths, moving to a rigid-flex construction) is confirmed case by case during review of the manufacturing files.

Two routing requirements follow directly from the copper choice:

• Route bend-zone traces perpendicular to the bend axis. Traces running parallel to the axis put one continuous strain direction along the whole conductor.
• No vias, no width transitions, no right-angle turns in the bend zone. A via is a structural discontinuity; a width step concentrates stress. The flex page states a 2 mil minimum trace width, but that is a capability limit, not a bend-zone recommendation — bend zones should use widths well clear of the capability edge.

Link 3: coverlay vs solder mask

Rigid boards use liquid photoimageable (LPI) solder mask. Flex bend zones cannot. Cured LPI is brittle; it cracks before the copper does, and the crack becomes the initiation site for copper failure underneath.

Flex uses coverlay instead: a pre-formed polyimide film laminated over the conductors with an adhesive, opened up only where pads must be soldered. The deployed flex page states the coverlay as PI/PSA — polyimide film with a pressure-sensitive adhesive system. The property that matters is material similarity: coverlay and base film are the same family, so both layers deform alike in a bend, with no "base still bending, surface already cracked" mismatch.

Three consequences to accept at design time:

1. Openings are mechanically formed, not photoimaged. Coverlay opening registration is not in the same class as LPI imaging, so pad-to-pad spacing requirements are looser than on rigid boards. Whether a fine-pitch area can be covered with coverlay, or needs a local alternative, is confirmed during manufacturing file review.
2. Coverlay adds total thickness. It is the most frequently omitted term in the ≥ 6× thickness bend radius calculation. Count the coverlay layers, then recompute the achievable radius.
3. No openings and no extra bonded layers in the bend zone. Keep the bend zone structurally continuous and push stiffeners, legends and openings outside it. That is the cheapest reliability decision available.

Hybrid schemes exist — solder mask on the non-bending assembly area, coverlay across the bend. This introduces a transition between two different surface layers; whether it is appropriate depends on the stack and the application, and requires engineering review.

Link 4: stiffeners write "no bending here" into the material

A flex circuit bends everywhere by default, and that is the problem. Connector mating areas, component areas and screw locations become failure sites if they move with the board. A stiffener's purpose is not reinforcement in general — it is a material-level declaration that this region does not participate in bending.

Three regions typically need one:

• Connector insertion areas. Mating is a repeated lateral load and the flex base cannot support it on its own; ZIF-type connectors additionally have thickness requirements at the insertion edge.
• Component mounting areas. The stiffness mismatch between a chip component and flexible base material is large. If the assembly area flexes, the solder joints take shear. The base must be rigid there.
• Fastener and mounting holes. Screw or clip clamping force needs something to bear against, or the area around the hole deforms.

Stiffener material, thickness and bonding method are project parameters. The flex page is explicit that material, stack-up, board thickness, copper weight, via type, surface finish and acceptance criteria are interdependent, and that manufacturing files must be submitted for project-level DFM review. So the selection-stage deliverable is these three inputs, with specifics confirmed by engineering review:

1. Stiffener location and outline — drawn on the drawing, not described in prose.
2. The function of the region (mating / assembly / fastening), because function sets the stiffness requirement.
3. Whether a transition distance to the bend zone is needed. A stiffener edge is a step change in stiffness; the bend should not start on that line.

That last point is the origin of the most common stiffener-related failure: placing the start of the bend exactly at the stiffener edge, concentrating all strain along it.

Link 5: dynamic flex vs form-and-stay

This is the question worth answering first, because it determines how the previous four links get decided. Both cases get called "flex", but the material paths are nearly separate.

Form-and-stay: the board is bent into position during assembly and never moves again. The classic case is replacing a wire harness between two rigid boards with a flex segment that is fixed once the housing closes. Material requirements are comparatively relaxed — adhesive-based construction is usable, ED copper is usable, copper weight has margin, coverlay follows normal practice. What must be respected is that the single forming event stays within the material's forming limit, and that no residual spring-back load ends up pushing on solder joints.

Dynamic flex: the board bends repeatedly through the product's service life. Clamshells, sliders, rotating joints, reciprocating print heads, wearable hinges. Here the choices tighten: prefer adhesiveless construction, RA copper in the bend zone, the thinnest workable copper weight in the bend zone, continuous coverlay with no openings, traces perpendicular to the bend axis, no vias in the bend.

The split also decides how acceptance works. Form-and-stay can be confirmed once after assembly for continuity and appearance. A dynamic bend-life requirement has to be written into project acceptance terms, with feasibility and verification method reviewed by engineering for the specific structure. The deployed flex page gives geometry (bend radius ≥ 6× board thickness) and material scope, not a bend cycle count — any specific dynamic bend life target must be confirmed in project documents and cannot be inferred from page specifications.

An actionable rule: if the product manual contains "the user will repeatedly open / rotate / extend" anything, select for dynamic flex; if the flex segment merely replaces internal wiring, form-and-stay selection is enough. The middle ground — infrequent, limited-count service bending — needs engineering review against cycle count and radius.

When to switch to rigid-flex

If you find yourself adding stiffener after stiffener to a pure flex circuit — assembly area, connector area, mounting holes — that is the signal that what you actually need is rigid-flex.

The structural starting point differs. A flex circuit is one flexible substrate where rigid regions are created by bonding stiffeners on top. A rigid-flex board integrates rigid and flexible layers during lamination, so the rigid regions are genuine rigid stacks and flexible layers extend out only in defined areas.

The deployed rigid-flex page states: 2–20 rigid layers, 1–8 flex layers, 3 mil minimum trace width, 0.3–4.0 mm board thickness, FR-4+PI base material. CLAIM_REGISTER records the rigid-flex layer ceiling as engineering-confirmed at up to 20 layers (conditional, requires engineering review). The page also lists "≥100,000 bend cycles" as a board-type specification; actual project bend requirements remain governed by project documents and confirmation records.

Three criteria:

• Rigid areas need multilayer routing — not just carrying components but routing and impedance control → rigid-flex. Asking 1–8 flex layers to serve both flexibility and rigid-area routing density gets tight fast.
• Many rigid-to-flex transitions at precise locations → rigid-flex. In rigid-flex the transition is a design object; on pure flex it is a by-product of where a stiffener edge happens to fall.
• Reducing connector count is a goal → rigid-flex. The rigid-flex page lists connector reduction as a defining benefit of the board type.

Conversely, if the board is one flex run with pads at both ends and the rigid area only has to hold a connector, flex plus stiffeners is the cheaper answer.

What to submit at selection stage

When material selection is done, these items determine whether engineering can return a manufacturability conclusion in one pass. Miss one and the review costs another round trip:

• Minimum bend radius and bend direction from mechanical (the input for capping total thickness);
• Use-case classification: dynamic flex or form-and-stay; for dynamic, a description of the motion;
• Bend zone extents: bend and non-bend boundaries marked on the drawing;
• Layer count and copper weight, annotated separately for bend and non-bend regions;
• Stiffener locations, functions and outlines;
• Coverlay opening locations, with confirmation that no openings fall inside a bend zone;
• Surface finish requirements (the flex page scope is that applicable finishes are reviewed against material, soldering, reliability and drawing requirements, with final options confirmed in project documents);
• Temperature and environmental conditions, particularly whether thermal cycling applies;
• Complete manufacturing files (Gerber, stack-up description, outline and bend definition drawing).

The flex page wording is worth pasting into the review email: material, stack-up, board thickness, copper weight, via type, surface finish and acceptance criteria are interdependent, and page content does not replace project confirmation. That is not a disclaimer — it says that changing any one of the nine items above moves the others, which is why they are submitted together.

To start a review, submit requirements from the Flex PCB page; to compare against an integrated construction see Rigid-Flex PCB; the polyimide grades in use with typical values are listed in the Materials Library; geometric and tolerance scope is on Manufacturing Tolerance; inspection and acceptance methods are on Quality Assurance; with files ready, go to Get Quote.

FAQ

How tight a bend radius can a flex circuit actually take?

The deployed flex page states bend radius ≥ 6× board thickness. It is a geometric constraint with total thickness as the variable, so the answer depends on your stack: base film, copper, adhesive and coverlay all add up. Getting tighter means getting thinner — adhesiveless construction, thinner copper, fewer layers — not changing a number. The achievable radius for a specific stack requires engineering review.

Is rolled-annealed copper always better than ED?

It is not a quality ranking, it is an application choice. RA copper's layered grain structure lies parallel to the bending strain direction, which suits repeated flexing. ED copper is entirely adequate where the copper does not bend, and it is the more standard option. The test is whether the board still moves after assembly: if it does, RA in the bend zone; if it does not, ED is usable.

Is adhesiveless material worth the extra cost?

It comes down to two questions: dynamic bending, and thermal cycling. If either applies, adhesiveless is worth evaluating — it removes the most dissimilar layer from the stack while making the assembly thinner, which relaxes the bend radius constraint. For purely static forming at ambient temperature on a cost-sensitive board, adhesive-based construction is sufficient.

Can solder mask replace coverlay on a flex circuit?

Not in the bend zone. Cured liquid photoimageable mask is brittle, cracks before the copper does, and the crack becomes a copper initiation site. The flex page states the coverlay as PI/PSA, the same material family as the base, so both deform alike. Hybrid schemes with solder mask on non-bending assembly areas exist, but they introduce a surface-layer transition and require engineering review.

Should I specify stiffeners, or is that the manufacturer's call?

Location, outline and function must come from the design side, because only you know where mating, assembly and fastening happen. Material, thickness and bonding method are manufacturability questions, confirmed by project-level engineering review after files are submitted. One thing to watch: do not place the start of the bend at a stiffener edge — that is a step change in stiffness.

When is pure flex the wrong choice?

Evaluate rigid-flex if any of these apply: rigid areas need multilayer routing with impedance control; there are many rigid-to-flex transitions requiring precise placement; reducing connector count is a goal. The deployed rigid-flex scope is 2–20 rigid layers, 1–8 flex layers, 0.3–4.0 mm thickness, FR-4+PI; the layer ceiling is engineering-confirmed at up to 20 layers and requires project review.