How Solar Panels Attach to a Roof, and Why Attachments Fail
By Seamless Home Team, Solar fulfillment operations · September 3, 2026
Quick answer
A rooftop PV solar array is held on by attachments that must engage the roof's structural framing — the rafters or truss top chords — rather than the sheathing, because sheathing has almost no withdrawal capacity and the uplift load on an array is a pull-out load. Each attachment is typically a lag screw or structural screw into the framing member, with a specified pilot hole and a minimum thread penetration into solid wood, carrying a rail that spans between attachments. How many attachments and how far apart is not a workmanship choice: it is set by a structural calculation based on the design wind uplift and snow load for that site and roof, the withdrawal capacity of the fastener in that wood species, and the rail manufacturer's allowable span. Waterproofing is a separate requirement solved by a flashing method appropriate to the roof covering — a flashing integrated into the shingle courses, a flashed mount, or a sealed deck-mounted base — rather than by sealant alone. Attachments fail for six recurring reasons: a fastener that missed the framing, insufficient thread penetration, over-torquing that strips the hole, spacing that exceeds what the structural calculation assumed, surface-sealed flashing laid over the covering instead of integrated into it, and mixing hardware outside a manufacturer's listed assembly.
An array weighing a couple of thousand pounds is held onto a house by perhaps thirty or forty lag screws, and kept from leaking by whatever was done at each of those thirty or forty holes. Both halves of that sentence are invisible within a day of the crew leaving, and both fail for reasons that are entirely predictable.
The structural half fails because a fastener did not engage what the calculation assumed it would. The waterproofing half fails because a penetration was sealed rather than flashed. Neither is exotic, and neither is discovered on the day.
The load path, which is the whole subject
Everything about attachments follows from one fact: the governing load case on a rooftop array is uplift, not weight.
Wind flowing over a roof produces suction, and an array presents a large surface for it to act on. The load that matters is therefore a load trying to pull each fastener straight out of the wood — a withdrawal load. Weight matters too, and snow matters in much of the country, but the case that sizes and spaces the attachments is usually the one pulling upwards.
The path runs: module frame → clamp → rail → attachment → fastener → framing member → structure. Each interface has a specification. Two of them carry almost all of the consequence, because they are the two nobody can see afterwards — the fastener's engagement with solid framing, and the flashing at that same hole.
Why sheathing is not an option
Roof sheathing is commonly half-inch plywood or oriented strand board, and its screw withdrawal capacity is a small fraction of a rafter's. A lag driven into sheathing alone can feel completely solid to the installer standing on it, because a person's weight is a downward load and the sheet is perfectly capable of resisting that.
Uplift is the opposite direction. The failure mode is not a dramatic one: it is a fastener withdrawing incrementally under repeated load cycles. This is why a missed rafter is not a defect that shows up immediately. It holds for years, and then does not, in the first storm that loads that corner properly.
There are engineered exceptions — certain structural insulated panels, metal decks, and proprietary systems with an evaluation report covering attachment to a defined deck assembly. Those are conditions defined by a manufacturer's evaluation, not general practice. On a conventional framed roof, an attachment that does not engage framing is a defect however solid it feels.
Spacing is an output of the design, not a field decision
There is no correct number of inches, and a general figure quoted on a roof is a warning sign. Attachment spacing is bounded by three independent limits, and the smallest governs.
| Limit | Set by | What changes it |
|---|---|---|
| Withdrawal capacity per attachment | Fastener diameter, thread penetration into solid wood, wood species | A different species or a shallower rafter changes the capacity materially |
| Design uplift per unit of array area | Site wind speed, exposure, building height, roof zone | Corners and edges see far higher uplift than the field of the roof |
| Rail allowable span | Rail manufacturer's span tables for that load | Snow load can make this the governing limit even when uplift sizes the attachments |
The roof-zone point is the one most often missed in the field. Uplift is not uniform across a roof: corner and edge zones carry substantially higher pressures than the middle. An array that reaches into a corner needs closer attachment spacing there, which means the spacing is not necessarily constant across one array — and a crew regularising it to a tidy uniform pitch has changed the design.
This is what the structural letter exists to state. It is issued against a specific layout, a specific attachment pattern and specific assumptions about the framing. Widening spacing in the field to avoid an awkward rafter does not just deviate from a drawing; it invalidates the document the permit was issued against, and it is the kind of change that turns into a design revision if it is caught and a latent structural problem if it is not.
Where the layout is also constrained by fire setbacks and access pathways, those two constraints interact: the module area available is bounded by fire code, and the attachment pattern within it is bounded by structure. Resolving one in the field without the other is how arrays end up either non-compliant or under-attached.
The fastener itself
Three specifications, all from the manufacturer's instructions rather than from habit:
- Thread penetration into solid framing. A minimum depth of thread engaged in the framing member, below the sheathing. This is the number the withdrawal capacity is calculated from.
- Pilot hole diameter. Undersized and the lag can split the rafter or shear on driving. Oversized and the threads never develop their capacity. Both are common, and the oversized case is worse because it looks fine.
- Torque. Enough to seat, not enough to strip. An over-torqued lag in a pilot hole that is slightly too large strips the wood threads, and a stripped attachment retains almost none of its design capacity while appearing fully installed.
Missing the framing edge is its own category. A fastener that catches the corner of a rafter has some engagement, is genuinely hard to detect afterwards, and delivers a fraction of the intended capacity.
Waterproofing: flashing, not sealant
Every attachment on a pitched shingle roof is a hole through the weather barrier, and the roof manages water by geometry — overlapping courses that direct water down and over, never relying on adhesion.
A compliant penetration works the same way. A flashing plate is integrated into the courses so water running down the roof passes over the flashing and off it, with the flashing under the course above and over the course below. It is the same principle as a plumbing vent flashing.
Methods that work:
- Integrated flashing plate woven into the shingle courses, the traditional and most durable approach on composition shingle.
- Flashed mount with a factory-integrated flashing, installed to the manufacturer's sequence.
- Deck-mounted base with a compressed butyl or EPDM seal against the deck, common with some proprietary systems — dependent on the seal being compressed exactly as specified, and on the hole not having been over-drilled.
- Tile replacement flashing or tile hooks on tile, where the tile sits back down over the detail.
- Seam clamps on standing-seam metal, which avoid penetration altogether.
- Roofer-executed detailing on low-slope membrane, which is generally not the solar crew's scope.
What does not work over time is a mount bedded in sealant on top of the covering. Sealant is a maintenance material with a service life measured in years, being asked to do a job the roof does with geometry and does for decades. It passes a hose test on the day of installation, which is exactly why it survives to become somebody's problem later.
The six ways attachments actually fail
- Missed framing — fastener in sheathing only, or catching a rafter edge.
- Insufficient thread penetration — the fastener is in the rafter but not far enough into it.
- Stripped or over-drilled hole — over-torqued, or a pilot hole too large, so the threads never develop capacity and a deck seal no longer compresses.
- Spacing beyond the calculation — regularised in the field, or widened to dodge a rafter, or uniform across a roof whose corner zone needed tighter spacing.
- Surface-sealed flashing — laid over the course rather than woven into it, or set too low relative to the penetration, or a fastener placed above the flashing's water line.
- Hardware mixed outside a listed assembly — a clamp or splice from another manufacturer, which is simultaneously a structural unknown and, because the same clamps commonly carry the bonding path, a grounding defect.
That sixth one is worth dwelling on, because it is the least intuitive. Mounting systems are evaluated as systems. The rail, clamps, splices and fasteners are assessed together, and the span tables, capacities and bonding path in the instructions apply to that combination. A substituted mid-clamp that fits the rail profile perfectly takes the installation outside the evaluation both the structural and the electrical compliance rest on. If a component is genuinely unavailable, that is a documented design change rather than a field decision.
What is visible afterwards, and what is not
This is the practical asymmetry that makes attachments worth attention before the modules go on.
| Defect | Detectable after completion? |
|---|---|
| Missed framing | Not without removing modules or specialist inspection |
| Thread penetration | No |
| Stripped hole | No |
| Spacing vs calculation | Partly — count and positions are measurable, the calculation's assumptions are not |
| Flashing sequence | Rarely — the detail is under the module and under the course above |
| Mixed hardware | Yes, if the system on the roof is compared against the plan set |
Five of the six are effectively unverifiable once the array is complete. That is what makes photographs during installation — of each attachment before the mount goes on, and of the flashing before the module covers it — the highest-value documentation on the whole project. It is also why some jurisdictions require a special inspection at attachment stage: somebody has to look while looking is still possible.
The upstream control is the site survey. Framing direction, spacing and member size, roof covering type and its remaining life, and layers of existing covering all belong in that visit, because each of them changes either the attachment design or whether the roof should receive an array at all. A survey that photographs the roof but not the framing has left the attachment design to be discovered on installation day, which is where the field decisions that invalidate calculations get made.
Roof age is the fact that decides the argument
Liability for workmanship sits with the licensed contractor whose permit the work was performed under. But disputes about roof leaks are almost never about responsibility — they are about causation, and causation turns on the condition of the covering when the array went on.
A leak at a penetration weeks after installation belongs to the installation. A leak three years later, near but not at a mount, on a covering that had eight years of life left when the array was installed, is genuinely contestable. Documenting roof condition before installation — photographs, covering type, apparent age, existing damage — is worth more than any argument afterwards, and it is the same document that answers whether the roof should have been replaced first.
The separate exposure is the roof covering manufacturer's warranty, which is a different instrument from the installation workmanship warranty and can respond differently to penetrations made by a party other than an approved installer. Which of the several warranties answers a given failure is a question worth resolving before it is needed rather than during a claim.
One practical consequence: attachments and flashing are the part of an installation that stops being visible once the modules are over them, which is why some jurisdictions inspect them mid-install and why covering them first is expensive.
The bottom line
Attachments are the least glamorous and least reversible part of a PV solar installation. Six defects account for nearly all of the failures, five of them cannot be verified once the modules are on, and the two consequences — an under-capacity array and a leaking roof — both arrive years after the crew has left.
The controls are unremarkable and they work. Establish framing location and roof condition at survey rather than on installation day. Treat the attachment pattern in the structural letter as a requirement rather than a suggestion, including tighter spacing in corner and edge zones. Photograph every attachment before the mount covers it and every flashing before the module covers it. And install the mounting system that is named on the plan set, complete, without substitutions.
Seamless Home coordinates design, structural documentation and permitting as part of design and permitting, which includes carrying the attachment pattern and the mounting system onto the plan set so the crew is building to a design rather than deciding on the roof. Installation is performed by vetted installing partners engaged as its subcontractors. Coverage is confirmed per service area rather than promised as blanket availability. If attachment corrections or roof-leak claims are recurring across your projects, get in touch and we can look at where they originate.
Frequently asked questions
What holds solar panels onto a roof?
On a typical pitched composition-shingle roof, a sequence of four things. An attachment — a lag screw or structural screw — driven into a rafter or truss top chord through the sheathing, which is what actually resists the load. A flashing or flashed mount at that penetration, which is what keeps water out. A mount or foot on top of that, carrying a rail. And rails spanning between attachments, to which the modules are clamped. The modules themselves are held by mid-clamps between neighbours and end-clamps at the ends of each run, torqued to the mounting system's specification. The load path runs from the module frame through the clamp into the rail, along the rail to the nearest attachments, through the fastener into the framing member, and from there into the building's structure. Every one of those interfaces has a specification, and the two that carry the most consequence are the fastener's engagement with solid framing and the flashing at the same point — because they are the two that are permanently hidden once the array is complete.
Do solar attachments have to hit a rafter?
On the ordinary framed roof, yes, and this is the single most consequential thing about a rooftop attachment. Roof sheathing — plywood or oriented strand board, commonly around half an inch — has very little screw withdrawal capacity, and the governing load case on an array is wind uplift, which is a withdrawal load pulling the fastener straight out. A lag in sheathing alone can feel entirely solid to the installer on the day and still be far below the capacity the structural calculation assumed. The failure mode is not a screw shearing; it is a fastener slowly withdrawing under repeated uplift cycles, which is why a missed rafter can hold for years and then let go in one storm. There are engineered exceptions — some structural insulated panels, some metal decks, and proprietary systems evaluated for attachment to sheathing or to a specific deck assembly — but those are cases where a manufacturer's evaluation report defines the condition, not cases of general practice. On a conventional roof, an attachment that does not engage framing is a defect regardless of how it feels.
How far apart should solar mounting attachments be?
Whatever the structural calculation for that specific roof produces, which is why a general number is not useful and a general number quoted on site is a warning sign. The spacing is bounded by three separate limits and the smallest one governs. The withdrawal capacity of the fastener in that wood species at that thread penetration sets how much uplift each attachment can carry. The design wind uplift for the site, exposure and roof zone sets how much load has to be carried per unit of array area — and roof zones matter a great deal here, because corners and edges see substantially higher uplift than the field of the roof, so an array reaching into a corner zone needs closer spacing there than in the middle. And the rail manufacturer's allowable span table sets how far the rail can bridge between supports for the module and load in question. Snow load enters as a downward case that can govern the rail span even where uplift governs the attachment count. The practical consequence is that spacing is an output of the design, and widening it in the field to avoid a difficult rafter invalidates the calculation the permit was issued against.
How are solar roof penetrations waterproofed?
By a flashing method matched to the roof covering, with sealant as a supplement rather than the primary defence. On composition shingles the durable methods integrate a flashing plate into the courses so that water running down the roof passes over the top of the flashing and off it, exactly as it does over a plumbing vent flashing — the flashing goes under the course above and over the course below. Alternatives that also work include a flashed mount with a factory-integrated flashing, and deck-mounted bases with a butyl or EPDM seal compressed against the deck under the mount, which are common with some proprietary systems and depend on the seal being compressed as specified. Tile roofs are handled differently, usually with a tile replacement flashing or a hook that allows the tile to sit back down over it. Standing-seam metal is often non-penetrating altogether, using clamps on the seam. Low-slope membrane roofs are generally the roofer's work rather than the solar crew's. What does not work over time on a shingle roof is a mount bedded in sealant on top of the covering: sealant is a maintenance material with a service life measured in years, and it is being asked to do a job the roof does with geometry.
Why do solar attachments leak?
Rarely because the sealant was the wrong brand, and almost always because the water management was wrong in principle or the flashing was placed wrong in sequence. The recurring causes are a flashing laid on top of the shingle course instead of woven into it, so that water runs onto the flashing's upper edge rather than over it; a flashing that is not high enough up the roof relative to the penetration; a nail or the lag itself placed above the flashing's water line; a fastener hole that was over-drilled or stripped, so the fastener no longer seals against the deck even where a gasket is present; and a penetration relocated in the field after the flashing was set, leaving the original hole to be patched with sealant. The second family of causes has nothing to do with the penetration: an array changes how water and debris move across a roof, so leaves collect behind rails and above mounts, holding water against the covering in places it was never designed to sit. That is a maintenance consequence rather than an installation defect, but it turns a marginal detail into a leak years later, which is where the disagreement about cause usually starts.
Can you mix solar mounting hardware from different manufacturers?
Not inside a listed assembly, and this catches out crews who are being careful rather than careless. A module mounting system is evaluated as a system: its clamps, rails, splices, attachments and fasteners are assessed together, and the allowable spans, capacities and bonding path in the instructions apply to that combination. Substituting a component from another manufacturer because it fits the rail profile and appears equivalent takes the installation outside the evaluation that the structural and electrical compliance both rest on. Structurally, the substituted part's capacity is unverified in that assembly. Electrically, it can break the bonding path, because the same clamps that hold the module are frequently the components that bond it, so a mechanically fine substitution can quietly become a grounding defect. The practical rule is that the mounting system named on the plan set is the system that gets installed, and if a component is genuinely unavailable, the substitution is a design change to be documented rather than a field decision.
Who is liable if a solar attachment damages the roof?
Responsibility for the workmanship sits with the licensed contractor whose permit the installation was performed under, but the practically difficult question is usually causation rather than responsibility. A leak appearing at a penetration shortly after installation is straightforwardly the installation's. A leak appearing three years later, on a roof that was already partway through its service life, at a location near but not at a mount, is where positions diverge — and where the age and condition of the covering at the time of installation becomes the decisive fact. This is why documenting roof condition before installation is worth more than any argument afterwards, and why a roof near the end of its life should be addressed before an array goes on rather than after. The related exposure is the roofing manufacturer's warranty on the covering itself, which is a separate instrument from the installation workmanship warranty and can respond differently to a penetration made by a party other than an approved installer.