Curtain Wall Engineering: Why a Facade Is a Movement Problem Before It Is a Strength Problem

A curtain wall carries no building load. It supports its own weight, resists wind pressure, and passes both back to the structure. In pure strength terms that is not a demanding brief, and the aluminium sections that do it are modest.

What makes facade engineering difficult is that this lightweight non-structural skin is attached to a heavy structure that moves. The floor slabs deflect under load and creep over years. The frame sways in wind and drifts in earthquakes. The facade expands and contracts through a daily and seasonal temperature range far wider than anything the interior structure experiences. And the glass in it is a brittle material whose strength depends on surface condition and load duration rather than on a fixed allowable stress.

This post covers what governs curtain wall design: accommodating structural movement, how glass strength actually works, the drainage principle that keeps water out, and the thermal behaviour that drives both performance and durability.

1. Movement Accommodation Is the Core Problem

A facade is attached to a structure at discrete anchor points, typically one or two per panel at each floor level. Between those points the structure moves and the facade must not be forced to move with it in a way that induces stress it cannot take.

Live load deflection and creep

A floor slab deflects when loaded. If a curtain wall panel spans from one floor to the next and is rigidly fixed at both, slab deflection above squeezes the panel vertically. Glass and rigid sealant joints cannot take that. The standard response is to support each panel's weight at one level only and connect it at the other with a slotted or sliding connection that permits vertical movement while still transferring wind load horizontally.

Concrete adds a second, slower movement. Creep and shrinkage continue for years after construction, producing additional deflection that can equal or exceed the initial elastic deflection. A facade installed on a young concrete frame will experience movement over its first several years that a facade on a mature steel frame will not, and the joint sizing has to allow for it.

Thermal movement

Aluminium has a coefficient of thermal expansion of roughly 13 microstrain per degree Fahrenheit, about twice that of steel and concrete. A dark-coloured aluminium mullion in direct sun can reach 160 degrees Fahrenheit or more, while the same member on a winter night sits below freezing. Over a temperature range of 150 degrees, a 20 foot mullion changes length by about half an inch.

That movement has to be accommodated within the system. Mullion splice joints between floors incorporate expansion gaps with slip connections. Where a facade runs continuously across a long elevation, thermal movement accumulates across the run and requires periodic movement joints in the same way a long building requires expansion joints.

The differential is as important as the absolute. The exterior face of a curtain wall moves with outdoor temperature while the interior face sits near room temperature, so the section bows. Thermally broken framing reduces this, but does not eliminate it, and the anchorage has to permit the resulting rotation.

13 µε/°F

Thermal expansion coefficient of aluminium, roughly twice that of steel or concrete. A 20-foot mullion cycling through 150°F changes length by about half an inch, which the system must absorb without stressing the glass.

Seismic drift

In an earthquake the structure racks laterally, and each floor displaces relative to the one below by the story drift. The facade attached across that interval has to accommodate the resulting shear distortion without the glass falling out.

Two approaches exist. Most conventional stick and unitised systems rely on glass rattle space, the clearance between the glass edge and the frame, which allows the frame to rack while the glass rotates within it rather than being loaded in shear. The available clearance sets the drift capacity, and it is verified by racking test to AAMA 501.4 or by calculation.

Unitised systems can go further by designing the stack joint between adjacent units as a sliding connection, allowing units to move relative to each other during drift so that each unit distorts less. This is the standard approach where drift demands are high.

The failure that matters is glass fallout, because falling glass from height is a life safety hazard well beyond the building itself. ASCE 7 requires that the facade accommodate a specified multiple of the design story drift, and for high occupancy or essential facilities the requirement is verified by dynamic racking test rather than by calculation alone.

2. Glass Is Not a Conventional Structural Material

Steel yields before it fails and gives warning. Glass does neither. It is perfectly elastic to failure, and that failure originates at surface flaws under tension.

Why surface condition governs strength

Theoretical glass strength is enormous, but real glass is covered in microscopic surface flaws introduced during manufacture, handling, and service. Under tension, stress concentrates at the tip of a flaw, and when the local stress exceeds the material's capacity the crack propagates instantly across the pane.

This makes glass strength statistical rather than deterministic. Two identical panes have different strengths because they have different flaw populations. Design therefore works on probability of breakage, conventionally 8 lites per 1,000 at the design load, rather than on an allowable stress with a safety factor.

It also makes edge condition critical. A cut edge has more and larger flaws than a polished one, and glass carrying bending has its highest tension at the edge. Edge treatment specification is a strength decision.

Load duration

Glass exhibits static fatigue. Under sustained tension, moisture at a flaw tip drives slow crack growth, so a pane that survives a 3-second wind gust may fail under the same stress applied for a month. Strength is therefore quoted against a reference duration, typically 3 seconds for wind, with reduction factors for longer durations. Snow, dead load on sloped glazing, and permanent loads on structural glass elements all require the long-duration values.

Heat treatment

Heat treatment introduces residual compression at the glass surface. Since failure requires tension at a surface flaw, applied tension must first overcome that compression, which raises the effective strength.

Annealed glass  has no residual stress. It is the baseline, roughly 6,000 psi allowable for short-duration load, and it breaks into large sharp shards.

Heat-strengthened glass  carries moderate residual compression and roughly double annealed strength. It breaks into large pieces that tend to stay in the frame, which is why it is preferred for laminated overhead and sloped glazing.

Fully tempered glass  carries high residual compression and roughly four times annealed strength. It fragments into small blunt pieces on failure, which is why it qualifies as safety glazing. The trade-off is nickel sulfide inclusion risk: a microscopic inclusion in the tension core can expand over years and trigger spontaneous breakage with no external cause. Heat soak testing screens for it by holding glass at elevated temperature to induce failure of susceptible units before installation.

Laminated glass

Two or more glass plies bonded with a polymer interlayer, usually PVB or a stiffer ionoplast. The interlayer holds fragments in place after breakage, which is what makes laminated glass the required construction for overhead glazing, guards, and structural glass elements where post-breakage retention matters.

Structurally, laminated glass sits between two bounds. If the interlayer transferred no shear, the plies would act independently and the assembly stiffness would be the sum of the individual plies. If it were rigid, they would act as a single thicker section, which is far stiffer. Real interlayers fall between, and their shear stiffness is strongly temperature and load duration dependent: stiff under short-duration cold conditions, much softer under sustained warm conditions. Design uses an effective thickness calculated for the relevant temperature and duration.

3. Water Control Works by Drainage, Not by Sealing

The instinct with a facade is to seal it against water. That approach fails, because sealant joints on a moving building eventually develop discontinuities, and once water gets past a single perfect barrier there is nothing behind it.

Modern curtain wall uses pressure-equalised rainscreen drainage instead. The system accepts that some water will pass the outer face and provides a drained cavity behind it to collect that water and return it to the exterior through weeps at each horizontal.

Why pressure equalisation matters

Water needs three things to cross a joint: water present, an opening, and a force to move it through. Rainscreen design attacks the third. Venting the cavity to the exterior lets cavity pressure equalise with external air pressure, removing the pressure differential that would otherwise drive water inward. Any water that still enters drains down the cavity and out.

This makes the inner line, the air and vapour barrier, the critical seal, and the outer face a shedding surface rather than a barrier. Continuity of that inner line, particularly at panel joints, corners, and transitions to adjacent construction, is what determines whether the system works.

Testing Is Part of the Design Process

Curtain wall performance is verified by test rather than by calculation alone. Laboratory mock-up testing follows a standard sequence: air infiltration to ASTM E283, static water penetration to E331, dynamic water penetration to AAMA 501.1, structural performance to E330, and seismic racking to AAMA 501.4 or 501.6. Field testing to AAMA 502 or 503 verifies installed performance on completed work. The mock-up is where detailing errors surface at a cost of a few panels rather than a whole elevation, which is why the mock-up programme belongs early in the schedule and why compressing it is usually a false economy.

4. Thermal Performance and Its Structural Consequences

Aluminium conducts heat about a thousand times better than the insulating glass it frames, so an untreated aluminium frame short-circuits the wall's thermal performance and becomes the coldest interior surface in winter.

Thermal breaks address this by interrupting the metal path with a low-conductivity polymer, either a poured and debridged polyurethane or a mechanically crimped polyamide strut. Structurally this means the mullion is two aluminium sections connected by a polymer, and composite action across that connection depends on the break material's stiffness, which varies with temperature. Design accounts for reduced composite action at elevated temperature.

Insulating glass units bring their own issue. The sealed cavity between plies contains gas at the pressure and temperature of manufacture. Change the temperature or the barometric pressure and that gas expands or contracts, deflecting the glass plies inward or outward. This produces visible distortion in reflections and, more importantly, stress at the edge seal. Units shipped over a mountain range or installed at an elevation well above the manufacturing plant can arrive with significant internal pressure differential, which is why capillary tubes are specified for units crossing significant elevation change.

Where interior surfaces run below the interior dew point, condensation forms. On a curtain wall the frame is the coldest surface, so it condenses first, and persistent condensation at the sill corrodes the frame and damages adjacent finishes. Condensation resistance is a specified performance metric, evaluated by AAMA 1503 test or by thermal modelling to NFRC procedures.

5. Where Facade Problems Actually Originate

Anchorage tolerance.  Structural frames are built to construction tolerances of an inch or more; facades are manufactured to fractions of an inch. The anchor is where that gap is closed, and it needs adjustability in all three axes plus the ability to accommodate the movements described above. Anchors detailed without adequate adjustment range are the most common source of field problems on facade installation.

Interface with adjacent construction.  The curtain wall system itself is engineered and tested. The junction where it meets a roof, a masonry wall, a spandrel panel, or a floor slab edge is frequently drawn at small scale and resolved in the field. Most water leakage on curtain wall buildings occurs at these transitions rather than within the tested system.

Sealant compatibility and joint design.  Structural silicone glazing relies entirely on the sealant bond, and bond depends on substrate preparation, primer selection, and compatibility with everything the sealant touches, including setting blocks, gaskets, and adjacent sealants. Incompatible materials can plasticise the silicone and destroy adhesion. Compatibility and adhesion testing on actual project substrates is a standard requirement for a reason.

Differential movement at corners.  Two elevations meeting at a corner move differently under thermal and structural loading, and the corner joint has to accommodate movement in two directions simultaneously. Corners are the most common location of glass breakage and sealant failure on completed facades.

Conclusion

Curtain wall design is dominated by movement rather than strength. The facade has to accommodate slab deflection and long-term concrete creep, its own thermal expansion at roughly twice the rate of the structure behind it, and story drift under wind and seismic loading, all without imposing stress on a brittle glazing material.

Glass behaves unlike conventional structural materials: strength is statistical and depends on surface flaws, it degrades under sustained load, and heat treatment changes both strength and failure mode. Water is controlled by drainage and pressure equalisation rather than sealing. And most facade failures originate not within the engineered system but at anchorage, at interfaces with adjacent construction, and at corners where two elevations move against each other.