Wood Floors Over Underfloor Heating: What the Substrate Has to Do

The complaint arrives in month four, usually in February. The floor was flat at handover, the client turned the heating up in December, and by the time the sun comes back there are hairline gaps at every third board and two planks that ride high at the ends. Nobody spilled anything. Nothing was installed wrong in any way an installer would recognise. The floor simply did what wood does when you put a heat source underneath it and nobody costed in the physics.

Choosing engineered wood flooring for underfloor heating substrate stability is not really a product decision. It is three decisions that happen at different times, usually to different people: which species and cut you buy, which construction the mill builds it in, and what the general contractor does to the screed in the eight weeks before the floor arrives. Get two right and one wrong and you still get the February call.

This piece covers all three, in the order they actually bite.

Movement is a species property before it is a construction property

Every hardwood expands and contracts across the grain as its moisture content changes. The rate is a material constant, published for most commercial species, and it varies by a factor of two or more between the species a designer might casually treat as interchangeable.

Two numbers matter. Radial shrinkage is movement across a quartersawn face. Tangential shrinkage is movement across a flatsawn face — roughly double the radial figure in most species, which is why cut direction is not a cosmetic choice. Published figures are green-to-ovendry totals; they are useful as a ranking, not as a spec value, because provenance and growth rate shift them.

Sources for these figures differ by a few tenths and none of them are a substitute for a supplier's own data, so treat the table as a ranking exercise. What it tells you is blunt: an oak floor and a teak floor over the same heating loop are not the same risk. If the design is fixed on oak, the mitigation has to come from cut, width and construction rather than from hope.

Two details that rarely make it into a specification but decide the outcome:

• The T/R ratio predicts cupping, not the tangential number alone. A board cups because its tangential and radial directions move by different amounts. White oak at a ratio near 1.9 will cup harder than walnut at 1.4 even though both are moderate movers. When a designer asks for "wide plank, flatsawn, over heating", the T/R ratio is the number to argue with.
• Width multiplies everything. Movement is a percentage of width. A 190 mm flatsawn oak board losing two points of moisture content moves on the order of a millimetre and a half; a 90 mm board in the same species and the same conditions moves half of that. Across twenty boards in a room, the first case is a visible gap pattern and the second is not. Halving plank width is the cheapest stability intervention available and it costs nothing at the mill.

Why an engineered construction is usually the right answer, even from a solid-wood mill

There is an awkward conversation that a good custom mill has with buyers and a bad one does not. Anrantabu Flooring is a custom handmade solid wood flooring maker in Huiyang, Huizhou — solid wood is the house speciality, and the four structural systems it builds across its range are solid, geothermal-compatible, three-layer and multi-layer. The geothermal-compatible line exists precisely because solid is the wrong answer over a heating loop most of the time, and a mill that will say so before you order is worth more than one that will sell you what you asked for.

The reason is structural, not commercial. A solid board is one piece of wood, so all of its movement is unopposed and all of it is across the width. An engineered board is a lamella of the show species bonded to a core whose grain runs across the lamella, plus a backing layer that balances the assembly. The cross-grain core physically restrains the lamella. The backing layer keeps the panel's moisture exchange symmetrical top-to-bottom, which is what stops it from curling like a leaf when one face is warmer and drier than the other.

That backing layer is the part buyers forget. A board with a 4 mm lamella on a core and no backing veneer is not a balanced construction. It is a bimetallic strip made of wood, and heat cycling is exactly the load case that reveals it.

Anrantabu builds the multi-layer engineered construction with anti-deformation substrate structures aimed at heating systems, including constructions that incorporate aluminium plate and plastic plate layers in the substrate. It also laminates its own substrates rather than buying board in, which matters for a reason unrelated to marketing: when the mill owns the lamination step, it also owns the moisture content of the core at glue-up, and core moisture at glue-up is the variable that determines whether the finished panel is flat six months later.

Ask about the construction in these terms:

• Is there a backing veneer, what species, and what thickness relative to the lamella?
• Does the core run perpendicular to the lamella across the full board, including under the tongue?
• Is the core solid softwood strip, plywood, or a composite with a plate layer — and what is the moisture content of the core at the moment of lamination?
• Is the adhesive rated for the temperature the board will see, and what is its formaldehyde class?

The surface-temperature ceiling, and the thermal resistance you have to budget

Two limits govern a heated wood floor, and they are set by different bodies for different reasons.

The comfort limit is a building-services number. EN 1264-2 caps floor surface temperature in occupied areas at 29 °C, with higher allowances in peripheral zones and bathrooms. The material limit is tighter: most European wood-flooring manufacturer instructions cap the wood surface at 27 °C, and the reason is not comfort but equilibrium moisture content. Wood held at 27 °C surface with normal room air sits at a much lower equilibrium moisture content than the same wood at 20 °C. Drive the surface hotter and you are running a slow kiln under the client's furniture.

The second limit is thermal resistance. EN 1264 works on the basis that the floor covering above the pipes contributes no more than about 0.15 m²K/W, because above that the system cannot deliver its design output without raising flow temperature past the surface cap. Wood conducts poorly — its thermal conductivity sits roughly in the 0.14 to 0.18 W/mK range depending on species and density — so the maths is unforgiving:

• A 14 mm engineered board is on the order of 0.09 m²K/W. Fine.
• A 20 mm solid board is on the order of 0.13 m²K/W. Almost the entire budget, before anything else.
• A 2 mm foam acoustic underlay can add 0.04 to 0.06 m²K/W on its own. That is what pushes a workable build-up over the line.

This is the calculation nobody runs until the system underperforms. If the floor is going over heating, ask the mill for the thermal resistance of the exact build-up you are buying — species, thickness, finish — and hand that number to the mechanical engineer before the loop spacing is fixed. A floor that arrives 4 mm thicker than the M&E drawing assumed is a commissioning problem, not a flooring problem, and it will be argued about for a month.

Related: the fixing method changes the answer. A floated floor over underlay always carries more resistance than the same board fully bonded to the screed, because the underlay and the air path are both insulators. Over heating, full-spread adhesive is usually the better thermal and dimensional choice, and it also removes the drum effect that clients complain about in floated installations.

The screed commissioning ramp, and why it is not the flooring supplier's problem until it is

More heated wood floors are wrecked by wet screed than by the wrong board. The sequence below is standard practice in the European market and worth writing into the contract even where local practice is looser, because it is the only part of the process the flooring supplier cannot control and will be blamed for anyway.

1. Let the screed cure before you heat it. Cement screeds are normally left at least 21 days before initial heating; calcium sulfate (anhydrite) screeds around 7 days. Heating earlier does not speed drying, it cracks the screed.
2. Run the functional heating cycle. Start at a flow temperature of roughly 20–25 °C and hold it for about three days. Then raise to the system's maximum design flow temperature and hold for at least four more days. This is a commissioning test of the loop, not a drying programme — the two are separate and both are needed.
3. Run the drying programme, then measure. After functional heating, cycle the system to drive out residual moisture, then test. Do not accept "it looks dry."
4. Test with a method, not a hand. In Europe the CM (carbide) method is standard: broadly, cement screed with heating is expected at or below about 1.8% CM, calcium sulfate with heating at or below roughly 0.3% CM. In the US, ASTM F2170 in-situ relative humidity probes are the accepted method, with acceptance thresholds set by the adhesive and flooring manufacturer, commonly in the 75–80% RH band. Record the readings, the locations and the date.
5. Wind the heat down before the floor goes in. Bring flow temperature down over several days and switch off roughly 48 hours before installation, so the slab is at room temperature when the boards are bonded.
6. Ramp back up slowly. After installation, wait — a week is common for adhesive cure — then raise flow temperature by no more than about 5 °C per day until you reach design. The first heating season is where a floor is won or lost.
7. Control the room, not just the floor. Target 18–22 °C air and 45–60% relative humidity through the heating season. Below 35% RH, gapping in any species is a matter of arithmetic, and no construction on earth prevents it.

One detail worth knowing: a heated floor that is going to gap will do it in its first winter, and a floor that is going to cup will do it in its first humid summer. Retention terms that expire at handover are retention terms that expire before the evidence arrives. If you are the importer, hold a portion of payment or a defect window across one full heating and one full cooling season, and say so at quotation stage — good mills price that risk in rather than argue about it later. This is the same logic that governs what documents to ask a China supplier for before the first order: the paperwork is only useful if it arrives before the money does.

What to ask the supplier next

The gaps in most heated-floor specifications are the same gaps every time. Send this list, in writing, and compare the answers across suppliers rather than reading any one of them in isolation. Anrantabu Flooring — a 20,000 m² plant in Yaohua Industrial Park, Huiyang, running timber drying, substrate lamination, cutting, assembly, hand-sanding and hand-rubbed wood wax oil finishing in-house — is one place to run the list; the questions work on any mill.

• Construction. Which of your structural systems do you recommend for hydronic underfloor heating, and why that one for my species and plank width? Is there a backing veneer, and what is it?
• Dimensions. What board thickness and plank width ranges do you offer in that construction, and what wear-layer thickness comes with each? (These are not published — get them in writing, per configuration.)
• Moisture. What is the target moisture content at despatch, and what tolerance band do you hold? What destination climate is that number set for?
• Tolerances. Send the dimensional tolerance table: thickness, width, length, squareness, and permitted cupping and bow, in millimetres. If a supplier cannot produce one, that is your answer.
• Thermal. What is the thermal resistance in m²K/W of the exact build-up quoted, and how was it derived?
• Temperature. What maximum wood surface temperature will you stand behind, and what happens to the warranty above it?
• Evidence. If you cite a third-party test report or a patented substrate structure, give me the report or registration number, the issuing body, the date, the exact sample or claim covered, and the current holder of record. Then verify it yourself at the source rather than accepting a PDF — the same discipline described in how to read a China certificate of conformity applies to test reports and IP filings alike.
• Samples. Custom samples are available on request; ask for the sample in the exact construction and thickness you intend to buy, not a showroom board.
• Commercial. Lead time is set per project schedule and payment is by T/T on terms agreed per order, so pin both to the commissioning calendar above, not to a nominal ship date.

Write all of it into the specification document rather than the email thread — a spec sheet a Chinese factory can actually quote from is the difference between comparable quotes and three suppliers answering three different questions. For the pattern work that sits on top of these constructions, the same thinking applies to how a Versailles parquet panel is built and to the material yield difference between herringbone and chevron. And if you want to see the constructions side by side before you commit, request them from Anrantabu Flooring as a single sample set — solid, three-layer and multi-layer in the same species — and put all three on a warm plate for a fortnight before you decide.