PVC roofing systems

The usual claim is that PVC resists grease and the others do not. The published trade guidance says something more precise and more useful, and the difference matters if you are going to defend a specification.

What is documented is a vulnerability, not a superiority. NRCA trade guidance records that EPDM membranes show vulnerability to animal fats and oils from vegetables and animals, as well as petroleum-based products such as gasoline and other fuels. That is what removes EPDM from the list in the deposition area. It is an affirmative finding about one material, not a clearance for another.

About PVC, the same source says: according to manufacturer information, PVC exhibits similar strengths and weaknesses in chemical compatibility as TPO — and that grease, diesels, fuel including jet fuels, solvents and oils should be avoided. Read that sentence twice. The membrane most commonly sold as the grease answer appears, in the trade literature, on a list of membranes for which grease is an exposure to avoid.

So why is PVC still the right place to start?

Three reasons, and they are worth separating because each one can fail independently.

1. The alternative is affirmatively documented as vulnerable. A specifier choosing EPDM for a restaurant roof is choosing against a published finding. That is a defensible reason to exclude it and a poor reason to feel confident about what replaces it.
2. The formulations engineered hardest against fats and oils sit inside this family. GAF literature, as reported in that same article, suggests its KEE-PVC blend undergoes less absorption and degradation when exposed to fats, oils, fuels, diesels and diluted acids than its other thermoplastic membranes. If you want the chemistry that has been pushed furthest in this direction, you are looking at KEE and PVC-KEE.
3. The system stays repairable. A grease-exposed roof accumulates damage at penetrations, at the exhaust curb, and wherever service technicians walk. On a hot-air welded membrane each of those is repaired by fusing new sheet to old, with no dependency on a tape system still being made and still being compatible. This page publishes no figure for how many years that lasts. It is not a chemical argument, but on this building type it is the argument that pays.

The honest position, then, is this: PVC and its KEE variants are the correct family to specify near commercial cooking exhaust, and the category name is not the specification. The chart for the actual product, checked against the actual exhaust stream, with the manufacturer’s written acceptance, is.

A resistance chart is a grid: substances down one side, a rating across the other — resistant, limited, not recommended, or some house equivalent. It looks like a specification. It is a summary, and the trade guidance that publishes comparisons of these charts says so in as many words: the resistance descriptions are not standard for all EPDM, PVC and TPO membranes manufactured but are based on themes present across manufacturers of the same membrane type, and it remains important to consult the manufacturer to be sure, on a case-by-case basis, how well a membrane will perform in a given situation.

The table below is this page’s own construction. It is not reproduced from any manufacturer document; it is the set of questions a chart cannot answer, matched to what to send back in writing.

One manufacturer publishes a list of roofing materials that create incompatible substrates for its PVC and PVC/CPA membranes, and suggests using a slip sheet or cover board between the new membrane and every item on it. The list is worth reading in full, because it is longer and more ordinary than most people expect: acrylic coatings; extruded polystyrene; polymer-modified bitumen; shingles; aluminum-coated asphalt; granulated cap sheets; that manufacturer’s own older roofs; TPO membranes; coated or smooth asphalt; polyurethane; sprayed urethane foam; expanded polystyrene; mineral-surfaced caps; coal-tar pitch; and PVC/CPA membranes.

Another manufacturer names the same mechanism more compactly — expanded polystyrene, extruded polystyrene, rubberized asphalt in some air and vapor barriers, asphalt, and old PVC roofs — and describes the remedy as undemanding: a simple bond break is all that is needed, for example polyethylene, a metal plate, insulation, fleece, or any other type of separator. Where the neighbour is a peel-and-stick rubberized-asphalt air barrier, its published answer is to separate the membranes using metal or a compatible sealant, because the rubberized asphalt can pull the plasticizers out of the flexible PVC and, in the other direction, those plasticizers can turn the rubberized asphalt to a liquid that runs down the membrane.

Two things about that list deserve an owner’s attention. Both are about procurement rather than chemistry.

It is a recover problem before it is anything else

Coated or smooth asphalt, polymer-modified bitumen, granulated cap sheets, mineral caps, coal-tar pitch — that is a description of most existing low-slope roofs in the United States. Any recover that leaves an asphalt system in place and welds a PVC sheet over it has to answer this question, and the answer belongs in the drawings and the scope of work. A proposal that is silent on it has not decided the question; it has deferred it to whoever is on the roof that morning.

The neighbours are on the list too

Sprayed urethane foam, polyurethane, TPO membranes, old PVC roofs. On a phased re-roof, on a building where two systems meet at an expansion joint, or where a coating was applied to part of the roof at some point in its history, the sheet is going to meet one of these. The tie-in detail is where that gets resolved, and it is the detail most often drawn as a cloud on a plan rather than a section.

One note on the source hierarchy, because it matters here more than anywhere else on this page: both lists are manufacturer technical bulletins, which is tier 6 evidence, product-specific. Neither is a standard and neither speaks for anyone else’s sheet. They are used here because the mechanism they describe is independently supported by peer-reviewed work, and because the party with the strongest commercial interest in this being simple is telling you it is not.

Thermoplastic membranes most often are seamed by heat-welding through the application of hot-air welding, and NRCA trade guidance gives the reason in one clause: heat-welding allows the material to bind to itself. That is the whole of the mechanical claim, and it is enough. This page publishes no history of the seaming methods that preceded it, because no source at an acceptable tier was verified for one.

The reason it matters commercially is repair. A welded lap is not a joint between two materials with a third one between them; it is one material. A patch welded onto that roof in year twenty is the same operation as the original seam, using the same equipment, with no dependency on a tape system still being manufactured and still being compatible. For an owner on a long hold, that is the difference between a maintainable asset and a countdown.

What conditions the repair

Three things, and a proposal should address all of them.

• The membrane has to be clean. One manufacturer’s welding instruction is unambiguous — always clean aged or dirty membrane before welding, with the cleaner specified for that membrane type. Published trade practice for aged PVC is a scrub with detergent and water containing trisodium phosphate, then a solvent wipe with MEK or acetone, then the weld. On a grease-exposed roof, that step is not optional and it is not quick.
• The parameters are product-specific. PVC runs hotter and slower than TPO. One manufacturer’s automatic welder settings list 1,094°F at 8.5 feet per minute for its PVC and KEE sheets against 1,004°F at 12.5 feet per minute for its TPO, and note that ambient temperature and sun versus shade move both. A peer-reviewed study that mapped the welding window for membranes from five manufacturers — two PVC, two TPO and one KEE — as received, after four days immersed in water, and after thirty days of severe soiling, and assessed the finished welds by peel testing, concluded by demonstrating the need for clear, product-specific welding guidelines for both new materials and for roof membranes exposed to the elements. Field conditioning changes the answer.
• The sheet has to be sound enough to weld to. An embrittled membrane may not be. That is the point at which the repairability advantage runs out, and it is the point this page’s next section is about.

The quality-control sentence worth memorising

Probing every seam with a blunt tool is the trade’s standard check and it belongs in every specification. It is also not enough. Two authors writing for the 2017 RCI International Convention — both employees of a membrane manufacturer, which makes the statement more interesting rather than less — put it directly: probing should be considered an insufficient measure to locate cold welds. A cold weld is a lap that looks finished and is not fused, and it can survive a probe and fail under thermal cycling two seasons later.

So ask what else is done. Test welds at the start of each shift and on each machine, before production seaming. Recorded temperature and speed settings for the conditions of the day. Destructive peel tests on cut samples, at a stated frequency, with the results in the closeout package. None of that is exotic; all of it is the difference between a seam program and a habit. See seam probe in the glossary.

What happened, and when

Early PVC membranes were manufactured without reinforcing materials and experienced numerous problems, the most consistent being that the membrane was prone to what the trade called shattering — substantial splits rather than local tears. In 1990 the National Roofing Contractors Association issued a bulletin to its member contractors alerting them of the shattering of aged unreinforced PVC roof membranes. Two years later, the same trade column records, NRCA and SPRI together warned members to avoid foot traffic on PVC roofs when the ambient outside temperature was below 50 degrees Fahrenheit — which tells you what an embrittled sheet in cold weather was understood to do. This page does not publish a failure rate, a population count or a thickness profile for those roofs; no source at an acceptable tier was verified for any of them.

Two things changed in response. Reinforcement — polyester or fiberglass — was added to all membrane sheets, which significantly reduced shattering by limiting how far a crack can run. And the plasticizers currently used are significantly better than the initial type, resulting in less plasticizer loss. Both are real improvements. Neither removes the mechanism, because the mechanism is intrinsic to a sheet whose flexibility depends on a mobile additive.

What a modern roof looks like when it goes

The 2022 study in Polymers is the most useful recent evidence, and it is worth being exact about what it is and is not. Researchers took three specimens from different segments of one flat roof after eleven years of service, because the waterproofing had degraded to the point that it no longer guaranteed watertightness. One specimen had been in contact with expanded polystyrene and exposed to sun; one had been in contact with expanded polystyrene and not exposed; one had been separated from the expanded polystyrene by a layer of polyester felt and not exposed.

The specimen protected by the polyester felt showed the least degradation due to dehydrochlorination; the exposed specimen in direct contact with the foam showed the most. Average surface chlorine measured by energy dispersive spectroscopy was lowest in that worst specimen at 21.58 percent by mass and highest in the separated one at 38.45 percent. Say the confounder out loud, because the paper does: the separated specimen was also shielded from direct weathering, by the felt and by gravel. So the felt is not isolated as the single variable. That is three specimens on one roof in one climate: evidence about a mechanism, and consistent with the value of a separator, but not a controlled trial of one, not a service-life figure and not a failure rate. This page treats it as exactly that.

The finding that should change how you write a warranty clause

The study’s abstract contains a sentence that has nothing to do with chemistry and everything to do with money. The mechanical properties of the degraded membranes with a polyester mesh backing, which are prescribed by the standards, do not usually deviate from the declared properties — and the authors note this often poses a problem for liability and warranty claims due to the poor quality of the waterproofing membrane.

Read plainly: a PVC roof can stop keeping water out while the sheet still passes the tests the standard specifies. If your dispute strategy is to send a sample to a lab and show that it fails the standard, that strategy may not work, and the authors are telling you why. Whatever else a commercial roofing warranty says, it should not be the only mechanism you are relying on. Ask what the warranty actually covers before you need to know.

The one thing an owner can still control

Not the formulation — that is chosen by a chemist you will never meet, and the record for any given product belongs to a version of it that may since have changed. What you can control is the separator. It costs almost nothing, it is fully within a specification writer’s gift, and the only peer-reviewed comparison this page could find of the same membrane with and without one points the same direction.

KEE is a separate standard, not a grade of PVC

Ketone ethylene ester has its own ASTM product standard, D6754. Products complying with it are internally reinforced with fabric and consist of polymers and other compounding ingredients with the KEE polymer being a minimum 50 percent by weight of the polymer content of the sheet. The 2024 IBC’s Table 1507.12.2 — model text — lists KEE against ASTM D6754 as its own row, separate from PVC and PVC/KEE against ASTM D4434. On the paperwork, then, KEE is a distinct material.

One practical wrinkle: the two standards do not use the same base thickness. D4434’s Types II and III physical requirements are based on minimum 45-mil sheets and Type IV on 36 mil; D6754 is based on a minimum 32-mil sheet. Published property values for a PVC product and a KEE product are therefore measured on different footings, which is worth knowing before you lay two spec sheets side by side and compare numbers.

PVC-KEE is a spectrum, and the threshold is where it turns

Below that 50 percent polymer-content threshold, a PVC-KEE alloy product typically complies with ASTM D4434 rather than D6754, and typically has some KEE-like physical properties. That is an honest description of a real middle ground — and it is also precisely the space where “KEE technology” language does the most work in a sales conversation. The question that settles it is not marketing: which ASTM designation does this product carry, and what are its published property values.

On the chemical side, the case for the KEE family is specific rather than general. GAF literature, as reported by NRCA trade guidance, suggests its self-produced KEE-PVC blend undergoes less absorption and degradation when exposed to fats, oils, fuels, diesels and diluted acids than its other thermoplastic membranes. That is a manufacturer statement about its own product line, which is the right way to weigh it: strong evidence that this is the direction the chemistry moves, weak evidence about any other manufacturer’s sheet.

CPA — what this page will and will not say

CPA stands for copolymer alloy. The only place it appears in the sources this page verified is inside one manufacturer’s own product designation — its PVC/CPA membranes — and inside that manufacturer’s incompatibility list, which names PVC/CPA membranes both as the product being installed and as one of the substrates requiring a slip sheet.

No standard-level source describing CPA as a generic material category was verified for this page. So the accurate statement is a narrow one: CPA is a designation used by at least one manufacturer for its own thermoplastic alloy, it is handled by that manufacturer under the same incompatibility rules as its PVC, and if it appears on a bid, the question to ask is the same one as for everything else — which ASTM standard does it comply with, and what are the published property values. This page does not describe it as a tier above or below PVC, because it found no basis for either claim.

On a re-roof, you already possess the best available map of where grease lands: the existing roof. The deposition pattern around an exhaust termination is visible as a darkened, tacky film, roughly elliptical, offset downwind of the fan and skewed toward the low point. Nobody has to model it. It has to be measured on the roof — by the consultant or contractor already carrying out the condition survey under their own fall-protection program, not by facilities staff sent up to look — and that measurement belongs in the condition report, alongside the core cuts and moisture survey.

This page does not publish a deposition radius, because no source at an acceptable tier supports one and the honest answer is that it depends on the hood, the volume, the filtration, the fan, the prevailing wind and the roof slope. What the arithmetic below does is show why measuring it changes the shape of the decision.

The arithmetic

Treat the affected area as a circle of radius r around the termination — a simplification, since the real pattern is offset downwind, but a conservative one if you measure r to the far edge of the visible film.

affected area = π × r²

On a 60,000 square foot roof with one exhaust termination:

• r = 15 ft → 707 sq ft → 1.2% of the roof
• r = 25 ft → 1,963 sq ft → 3.3% of the roof
• r = 40 ft → 5,027 sq ft → 8.4% of the roof

Even at a generous forty-foot radius, the chemically constrained area is under a tenth of the roof. That looks, at first, like an argument for treating the grease zone separately and specifying something cheaper everywhere else.

Why the obvious conclusion is wrong

Because of one sentence from a membrane manufacturer’s own technical bulletin: TPO and PVC are not interchangeable and will not weld to each other. A two-membrane roof cannot be seamed together. Tying the systems together requires a positive break between them and a separately waterproofed transition — a permanent, non-weldable joint running through the field of the roof, in a location chosen by where the grease happened to land.

Which gives a decision rule that is actually usable:

1. One membrane over the whole roof, from the PVC and KEE family. No transition joint, one weld procedure, one repair material, one set of spare stock. On most single-building roofs this is the right answer even though most of the roof does not need the chemistry, because the cost of the joint you avoid is larger than the premium you pay.
2. Two membranes only where a real boundary already exists. An expansion joint, a firewall, a genuinely separate roof area with its own drainage, or a phasing line that follows a structural break. If the boundary is a building feature rather than a convenience, it can be detailed properly and inspected later.
3. A sacrificial layer, not a second membrane. Where the deposition zone is small and well defined, some designs protect the membrane in that zone with a sacrificial surface or a containment system, so the damage accrues to something replaceable. Whether any given product is compatible with the membrane under it is exactly the substrate question this page keeps returning to, and it is answered by the membrane manufacturer in writing, not by the product’s own literature.

And underneath all three, the cheapest intervention of the lot, which is not a roofing decision at all: reduce what lands. A grease-collection device at the exhaust termination is a mechanical-code item, maintained by whoever maintains the kitchen exhaust. It removes the substance before the membrane ever sees it. On a building where the roofing scope and the mechanical scope are let separately, it is also the item that most reliably falls between them.

Two limits on this example, stated plainly. The circle is a simplification; the real deposition pattern is offset and irregular, and the measurement replaces the model. And the percentages are arithmetic, not evidence — they show how the constrained fraction scales with radius on a roof of that size, and nothing else.