The mechanical and thermodynamic estimates of Lesson 1.2 place the theoretical tensile strength of water near −103 atm. The actual values measured in real experiments span three orders of magnitude below that, depending entirely on what the experimenter does to clean and prepare the sample.
sample:
Pull the upper piston outward. The trapped water responds elastically — its bulk modulus K ≈ 2.2 GPa makes it nearly as stiff as a soft metal — so a small volume strain produces a large negative pressure. If the liquid were homogeneous all the way to its theoretical cohesive limit (green dashed, ≈ −1000 atm) the elastic line would continue to about 5% strain before any rupture. In practice every real sample contains *nucleation sites* — dust particles, gas microbubbles trapped in crevices, dissolved-gas concentrations — and the liquid tears at whichever site is weakest. Tap water tears at fractions of an atmosphere of tension; a meticulously prepared inclusion-free capillary preparation (Briggs 1950) holds about 270 atm. The gap between theoretical and measured tensile strength is the subject of the next chapter.
A century of tensile-strength experiments
The first systematic measurements of liquid tensile strength were Marcellin Berthelot’s capillary-tube experiments in 1850. Berthelot sealed water in a thick-walled glass capillary at elevated temperature; as the tube cooled, thermal contraction of the water (which contracts faster than the glass) put the water under tension. The water held a few atmospheres of negative pressure before snapping — orders of magnitude short of the theoretical limit, but enough to demonstrate that liquid water can be put into tension at all.
A century of refinement followed. The pattern, across roughly two hundred independent measurements, is consistent:
Sample preparation
Tensile strength held
Tap water
~−0.1 to −1 atm
Distilled water
~−1 to −10 atm
Distilled + degassed
~−10 to −100 atm
Inclusion-free quartz capillary (Briggs 1950)
~−270 atm
Micron-sized water droplet (Zheng et al. 1991)
~−1400 atm
The trend is unambiguous: the cleaner and smaller the sample, the higher the tensile strength it holds. Tap water tears almost instantly under any tension. Distilled and degassed water holds tens of atmospheres. A water sample prepared in an inclusion-free clean glass capillary holds a few hundred atmospheres. A micron-sized water droplet — small enough that the entire droplet is statistically unlikely to contain even a single nucleation defect — holds roughly the theoretical limit.
The dependence on sample cleanliness tells us, with overwhelming clarity, that the bulk-water cohesive limit is not what fails in any normal-sized sample. What fails is a nucleation site — a localised weakness in or on the sample whose threshold sets the tensile strength of the whole.
The Briggs capillary experiment
L. J. Briggs’s 1950 experiments at the National Bureau of Standards are the canonical measurement of the inclusion-free limit. Briggs sealed degassed, filtered water in a spinning glass capillary, the rotation of which produced a centrifugal pressure gradient along the capillary’s length. The pressure at the centre of rotation could be made arbitrarily negative by spinning faster; the meniscus position revealed the moment of cavitation. Briggs achieved tensions of −264 atm at 10 °C — three orders of magnitude above tap water, three times below the theoretical limit.
Briggs’s apparatus is to this day approximately the cleanest practical method of putting bulk water under tension; modern experiments using inclusions-free fluid inclusions in quartz crystals (Zheng et al. 1991) have pushed the limit closer to the theoretical value, but those measurements are on microscopic samples and require special preparation techniques.
⏳The history— The early history of liquid tensile strength
The first hint that liquids could be put into tension came from Christiaan Huygens in 1660, who noted that mercury in a sealed Torricelli barometer could be inverted without separating from the closed end — the mercury was held up against gravity by tension, requiring no airspace gap. The observation went unexplained until the molecular theory of matter matured.
Berthelot’s 1850 capillary measurements were the first quantitative tensile-strength data. Reynolds in 1873 made the connection to cavitation explicit: he observed that water flowing through a constriction below atmospheric pressure could cavitate — produce vapour bubbles — and that this phenomenon was distinct from boiling. Reynolds was also the first to identify cavitation as the cause of mysterious noise and surface erosion in early ship propellers; the propeller fully entered cavitation physics in 1893 with the inquiry into the lacklustre performance of the British battleship HMS Daring.
Briggs’s 1950 paper Limiting negative pressure of water in the Journal of Applied Physics is the standard data point for the inclusion-free tensile strength of pure water. The Zheng / Henderson 1991 measurements in microscopic fluid inclusions in synthetic quartz pushed the achievable tension to within a factor of two of the theoretical limit and remain the strongest experimental support for the bulk cohesive-limit estimates.
The shape of the next chapter
The interactive above shows what an idealised experiment looks like. Drag the strain up — the liquid responds elastically along a steep line whose slope is the bulk modulus — until the heterogeneous threshold is reached, at which point a vapour bubble nucleates and the tension immediately drops back to the threshold value. The threshold itself, set by whichever nucleation site is weakest, is what the next chapter computes.
Two paths are open. Homogeneous nucleation asks for the rate at which thermal fluctuations alone produce a vapour bubble larger than a critical radius — the answer involves the Boltzmann factor of a Gibbs free-energy barrier, and gives the right tensile strength only for samples that genuinely contain no other defects (cleaning is then sufficient). Heterogeneous nucleation asks how preexisting defects — dissolved gas, particles, gas pockets trapped in surface crevices — lower the barrier locally and provide preferred sites for the bubble to nucleate.
For any normal sample, heterogeneous nucleation dominates. The cavitation thresholds we measure are not the bulk water’s; they are the thresholds of the weakest defect in the sample. This is the puzzle the next chapter resolves.