Fire Sprinkler Head: The Glass Bulb That Bursts at 68 °C

Anatomy: a brass valve held shut by a glass strut

A standard pendent sprinkler is a valve that is also its own heat detector. Its brass body screws into a fitting on a water-filled branch line (a K5.6 head uses a ½ in NPT thread) and carries five working parts:

• Orifice and seal. The waterway ends in a smooth orifice closed by a metal cap and, in most modern designs, a PTFE-coated Belleville spring: a dished washer that seals against the seat and stores spring energy.
• Frame. Two arms loop from the body to a boss directly over the orifice.
• Compression screw. Threaded through that boss and torqued at the factory, it jams the bulb against the cap and compresses the Belleville spring, preloading the whole joint.
• Glass bulb. A sealed ampoule, typically 3 mm or 5 mm in diameter, holding a dyed proprietary liquid and a small bubble. It alone holds back line pressure.
• Deflector. A slotted disc that turns the jet into spray. Upright and sidewall heads use differently shaped deflectors.

The bulb is a compression strut. Worked number: heads are commonly rated 175 psi (12 bar), and at that pressure the water force on a ½ in (12.7 mm) orifice is p × A = 1.21 MPa × 1.27 × 10⁻⁴ m² ≈ 150 N, the weight of a ~15 kg mass. The preload is set above that so the seal stays tight through surges. Glass is strong in compression and weak in tension, so the bulb carries this axial load for decades, then fails from the inside when heat pressurises it.

Step by step: from hot ceiling gas to a spray umbrella

• 1. Hot gas reaches the ceiling. The fire plume strikes the ceiling and spreads outward as a thin ceiling jet moving a few metres per second, right where the heads sit.
• 2. The bubble shrinks. The liquid expands roughly 0.2% of its volume per °C while the glass barely grows, so the bubble is squeezed. While it remains, it cushions the liquid and internal pressure stays modest.
• 3. The bubble vanishes. In a 68 °C Viking bulb it is gone at ~55 °C, ~13 °C before it bursts (Yates et al., 2011). The liquid is now trapped in a nearly rigid shell.
• 4. Pressure soars. Each further degree adds tens of bar, loading the thin wall in hoop tension until, at the 68 °C rating, the glass fails.
• 5. The glass shatters. Cracks in glass run at up to ~1,500 m/s, crossing a bulb ~2 cm long in roughly 13 µs. The compressed liquid and the strut load drive the fragments apart.
• 6. The seal blows out. With the strut gone, the Belleville spring snaps back and water pressure shoves cap and spring clear of the orifice.
• 7. The jet hits the deflector. At NFPA 13's minimum of 7 psi (0.5 bar) the jet leaves at ~10 m/s, strikes the deflector and breaks into droplets covering up to ~21 m² (225 ft²) per head in light-hazard spaces.

The head never recloses: water flows until the control valve is shut, and the fired head is replaced from the on-site spare-head cabinet NFPA 13 requires.

The governing physics: a bubble that sets the trigger

The bulb is a liquid-in-glass thermometer with nowhere to rise: expansion goes first into the bubble, then into the glass.

Bubble phase. The bubble must absorb the liquid's expansion from room temperature to the vanishing point, so its volume fraction is roughly β × ΔT. Worked: with β ≈ 0.002 per °C, a bubble that closes at 55 °C must fill about 0.002 × (55 − 20) ≈ 7% of the internal volume at 20 °C. The void's size sets the trigger point: a bigger bubble takes more heat to fill. Not every bulb loses its bubble: Yates et al. found air-filled ASCOA bulbs burst with a speck of void still left.

Pressure phase. With the bubble gone, a rigid-bulb estimate is Δp ≈ K·β·ΔT, where K is the liquid's bulk modulus. For an organic liquid K ≈ 1 GPa, so Δp/ΔT ≈ 10⁹ Pa × 0.002 ≈ 2 MPa, about 20 bar per °C. A real bulb flexes and the liquid's properties drift with temperature, but even a conservative 10 bar per °C adds over 100 bar across the ~13 °C between bubble-gone and burst. So the bubble absorbs everyday temperature swings at near-zero stress for decades, then the last stretch converts heat into pressure steeply enough for a sharp, repeatable burst.

The colour ladder. The liquid is dyed to show the rating: orange 57 °C, red 68 °C, yellow 79 °C, green 93 °C, blue 141 °C, with purple and black above. NFPA 13 caps the ceiling temperature for each class: 38 °C (100 °F) for ordinary 57 °C and 68 °C heads, 66 °C (150 °F) for intermediate 79 °C and 93 °C heads, and 107 °C (225 °F) for high-temperature 141 °C heads.

Response time: RTI and the 3 mm versus 5 mm bulb

A head bursts when the bulb reaches 68 °C, but glass and liquid take time to warm, so in a growing fire the surrounding gas is much hotter by then. Heskestad and Smith at Factory Mutual (1976) captured this lag in one number, the Response Time Index: RTI = τ√u, where τ = mc/(hA) is the element's thermal time constant and u the gas speed. Because the convective coefficient h grows roughly as √u, the product is nearly a property of the head alone, in (m·s)^½.

RTI comes from the plunge test: a room-temperature head is plunged into a tunnel of hot air at fixed temperature and speed, and its operating time t gives RTI = t√u / ln[ΔT_gas / (ΔT_gas − ΔT_rating)], both rises measured from ambient. Modern standards add a conduction factor, C, for heat lost to the frame and pipe.

Worked example. Plunge a 20 °C head with a 68 °C bulb into 150 °C air at 2 m/s. The bulb must climb 48 of the 130 °C available, so t = (RTI/√2) × ln(130/82) ≈ (RTI/1.41) × 0.46. At RTI 50 it operates in ~16 s; at RTI 100, ~33 s.

Bulb diameter d drives RTI. Stored heat scales with d², surface area with d, and convection improves on thinner cylinders (h roughly ∝ d^−½), so τ scales roughly as d^1.5: (3/5)^1.5 ≈ 0.46. That is why a 3 mm quick-response bulb reaches RTI ≤ 50 (m·s)^½ while a 5 mm standard bulb sits at RTI ≥ 80 (m·s)^½, matching NFPA 13's fast- and standard-response definitions. NFPA 13 generally requires quick-response or residential heads in light-hazard occupancies such as offices and hotels.

The water side: K-factor, 7 psi and the umbrella

Once open, a head is simply an orifice, and flow follows Q = K√P (Q in gpm, P in psi). The metric K, in L/min per bar^½, is about 14.4 times the US value, so a K5.6 head is a K80.

• At the minimum. NFPA 13 requires at least 7 psi (0.5 bar) at any head. A K5.6 head then flows 5.6 × √7 ≈ 14.8, so ~15 gpm (~56 L/min). The ideal jet speed is √(2Δp/ρ) = √(2 × 48,300 Pa ÷ 1,000 kg/m³) ≈ 9.8 m/s, about 10 m/s.
• At a design point. For Ordinary Hazard Group 1, NFPA 13's density/area method calls for 0.15 gpm/ft², and a head may cover up to 130 ft² (12.1 m²). That head needs 0.15 × 130 = 19.5 gpm, so P = (Q/K)² = (19.5 ÷ 5.6)² ≈ 12 psi (0.83 bar). Doubling flow takes four times the pressure, which is why warehouse storage uses bigger heads, from K11.2 to K25.2 and beyond (NFPA 13 now lists up to K33.6).

Standard-spray heads, introduced in the 1950s, throw nearly all their water downward in a hemispherical pattern; the old-style heads they replaced sent 40–60% up at the ceiling. NFPA 13 keeps standard-spray heads at least 6 ft (1.8 m) apart so one head's spray does not chill its neighbour's bulb (cold soldering), and requires 18 in (457 mm) of clearance below the deflector so stored goods do not block the umbrella.

Standards, testing and history

Heads are listed to UL 199 in the US, approved to FM Approval Standard 2000, or certified to ISO 6182-1 and EN 12259-1. They are installed to NFPA 13 (or EN 12845 in Europe) and maintained to NFPA 25. Key type tests:

• Operating temperature (bath) test. Heads are warmed slowly in a liquid bath. Under ISO 6182-1 a bulb must operate within ±(0.035X + 0.62) °C of its rating X. For a 68 °C bulb: 0.035 × 68 + 0.62 ≈ 3.0 °C, so between about 65 and 71 °C.
• Bulb strength. Samples are crushed and analysed statistically; the weakest expected bulb must carry about twice the highest expected assembly load.
• Thermal shock. A bulb heated to just below its rating and dunked in cold water must not crack.
• Plunge, leak, strength and functional tests. RTI is measured, bodies are pressurised well beyond their rating, and heads are operated across a wide pressure range to prove the cap and spring never lodge in the frame.

Henry Parmelee patented an automatic sprinkler in 1874 to protect his New Haven piano factory, Frederick Grinnell's improved 1881 design made it practical, and NFPA's first standard, in 1896, covered sprinkler installation. Early heads used fusible solder links; in the 1920s Grinnell's Quartzoid bulb, filled with a coloured liquid, brought the glass bulb. The case for sprinklers is written in fires without them: the 1980 MGM Grand fire in Las Vegas began in an unsprinklered restaurant and killed 85 people, and The Station nightclub fire in Rhode Island (2003) killed 100 in an unsprinklered building. Both drove new sprinkler requirements.

Failure modes, myths and look-alikes

• The water is off. In NFPA data the leading reason sprinklers fail to operate is that the system had been shut off; a closed control valve defeats every head.
• Seals that stick. In 1998 Central Sprinkler Company recalled some 8–10 million Omega heads whose rubber O-ring seals could stick and keep the head shut.
• Paint, dust and corrosion. Paint or grease insulates the bulb and can glue the cap. Under NFPA 25, painted, loaded, corroded or damaged heads, and bulbs that have lost liquid, are replaced, not cleaned; samples are lab-tested after 50 years in service (20 for fast-response heads).
• Accidental discharge. A knock breaks a bulb just as heat does: clothes hangers on hotel-room heads, forklift masts and gym balls cause floods, hence guards and concealed heads. Wet pipes in unheated spaces can freeze and burst, which is what dry-pipe systems are for.
• Wrong head or blocked spray. A standard-response head where quick-response is required, too high a rating, or storage stacked into the pattern all delay or weaken control.

Myths. Smoke, shower steam and burnt toast do not trigger a sprinkler; only heat at the bulb does. Nor do heads all fire together. That movie scene is a deluge system, whose open nozzles have no bulbs: a deluge valve, tripped by separate smoke, heat or flame detectors, floods every nozzle in the zone at once. It suits aircraft hangars and transformer yards, not offices.

Look-alikes. Fusible-link heads melt solder instead of bursting glass, and dry-pipe and pre-action systems still open head by head; a smoke alarm only makes noise. Head-by-head sprinklers work: NFPA data put the civilian death rate per fire 90% lower when sprinklers are present.

A glass-bulb sprinkler head versus the devices and systems it is confused with

Device or system	What triggers it	What discharges	Typical use
Glass-bulb sprinkler (wet pipe)	Liquid in the bulb pressurises and bursts the glass at its rating, e.g. 68 °C	Only the heads heated past their rating	Offices, hotels, homes, shops
Fusible-link sprinkler	A solder-alloy link melts and the lever strut falls apart	Only the heads heated past their rating	Concealed heads; some residential and storage heads
Dry-pipe or pre-action system	Closed heads on air-filled pipes; pre-action also needs a detector signal	Still only the heated heads	Unheated car parks, freezers, data centres, museums
Deluge system	Separate smoke, heat or flame detectors trip a deluge valve	Every open nozzle in the zone at once	Aircraft hangars, transformer yards, chemical plants
Smoke alarm	Smoke particles (photoelectric or ionisation sensor)	Nothing: it only sounds an alarm	Early warning in homes and buildings