Why a Thick Tumbler Cracks Before a Thin One

This page contains affiliate links. If you buy through one we may earn a commission, at no cost to you.

Because a thick wall keeps the outside cold for longer while the inside is already hot, and glass does not break from heat. It breaks from a difference in heat. The inner surface tries to expand, the cold outer skin refuses to let it, and the outer skin ends up in tension, which is the one kind of load glass is bad at.

The arithmetic is unusually satisfying here, because it predicts the exact thing everybody has experienced. Pouring boiling water into a room temperature glass applies an 80 kelvin step. A thin soda lime wall needs about 87 kelvin to fail. A thick one needs only about 62. So 80 falls between them, which is why thin glasses usually survive, thick ones usually crack, and the whole business feels like luck rather than physics. It is not luck. You are standing in the gap.

Strong in compression, weak in tension

Glass has no mechanism for yielding. A metal under too much load slides its crystal planes past each other and deforms permanently, which blunts cracks and absorbs energy. Glass is an amorphous solid with no planes to slide, so when the stress at the tip of a flaw exceeds what the bonds can take, the flaw simply runs.

That gives glass a strange pair of strengths. In compression it is extremely strong, several hundred megapascals or more, because squeezing a crack closed stops it propagating. In tension it is weak and, more importantly, inconsistent, because its tensile strength is not really a material property at all. It is set by the worst flaw on the surface. Pristine glass fibre can reach several gigapascals. A tumbler that has been through a dishwasher a hundred times and has microscopic scratches all over it is worth perhaps 30 to 50 megapascals. Everything on this page uses 40, and that choice is why your results will vary from your neighbour's with the same glass.

So the whole question becomes: when you pour, which part goes into tension? And the answer is the part that stays cold, which is the outside. The hot inner surface wants to expand and is in compression, where glass is strong. The cold outer surface is being stretched by the expanding material behind it, and that is where the crack starts. This is worth knowing because it is counterintuitive in a useful way: the damage begins on the surface that never touched the hot water.

Two numbers decide it, and one of them is the whole reason borosilicate exists

The stress a temperature difference generates in a constrained surface layer is the elastic modulus, times the thermal expansion coefficient, times the temperature difference, divided by one minus Poisson's ratio. Three of those four are material properties, so each glass has a fixed stress per kelvin.

Soda lime glass, which is ordinary window and tumbler glass: expansion coefficient 9.0 parts per million per kelvin, modulus 70 gigapascals, Poisson's ratio 0.22. That gives 70e9 times 9.0e-6 divided by 0.78, which is 808 kilopascals per kelvin. Call it 0.808 megapascals for every kelvin of difference across the wall.

Borosilicate glass: expansion coefficient 3.3 parts per million per kelvin, modulus 64 gigapascals, Poisson's ratio 0.20. That gives 64e9 times 3.3e-6 divided by 0.80, which is 264 kilopascals per kelvin, or 0.264 megapascals.

Divide one by the other: 0.808 over 0.264 is 3.06. Borosilicate develops just under a third of the stress for the same temperature difference, and almost all of that advantage is the expansion coefficient rather than the stiffness.

Now convert to the thing you can act on. At a working tensile strength of 40 megapascals, soda lime fails at 40 divided by 0.808, which is a 49.5 kelvin difference. Borosilicate fails at 40 divided by 0.264, which is 151 kelvin. And a boiling water pour into a room temperature glass is an 80 kelvin step. Fifty is less than eighty. A hundred and fifty is comfortably more. That single comparison is the entire reason laboratory glassware and ovenware are borosilicate, and it is a far better reason than any listing gives.

Where thickness actually comes in, and how much it is worth

Now the thickness question, carefully, because the standard explanation is directionally right and skips a step that matters.

The standard explanation is that a thick wall develops a steeper gradient so it sees more stress. The hole in it is this: for a genuinely instantaneous surface temperature step, the peak surface stress does not depend on thickness at all. The instant the surface jumps to 100 degrees, it is constrained by a bulk that is still at 20, and that is true whether the bulk is two millimetres thick or two hundred. So thickness cannot enter through an idealised step.

What makes it enter is that real heat transfer is not instantaneous. Hot water delivers heat to the glass surface at a finite rate, set by a heat transfer coefficient, and the glass carries it away inward at a rate set by its conductivity. The ratio of those two is the Biot number: the heat transfer coefficient times the thickness, divided by the thermal conductivity.

Glass conducts at about 1.0 watts per metre per kelvin. Free convection of near-boiling water against a solid runs somewhere around 1000 watts per square metre per kelvin. So for a 2 millimetre wall the Biot number is 1000 times 0.002 divided by 1.0, which is 2.0. For a 6 millimetre wall it is 6.0.

A low Biot number means the surface heats slowly compared with how fast the wall can even out, so the surface never gets far ahead of the bulk and the peak stress falls well short of the idealised value. A high Biot number means the surface heats fast compared with the wall's ability to even out, so the peak stress approaches the idealised value. Using the standard approximation that the peak stress fraction is the Biot number divided by the Biot number plus about 1.5, the thin wall reaches 0.571 of the ideal stress and the thick wall 0.800.

Divide those: 0.800 over 0.571 is 1.40. So the thick glass is about 40 percent worse, and the honest answer to the question is "yes, and by less than you would guess from how reliably it happens". Forty percent is enough, because of where the 80 kelvin pour sits. Dividing the 49.5 kelvin failure difference by each fraction, the thin wall needs an applied step of 86.7 kelvin to fail and the thick wall needs 61.9. Eighty is between them. A 40 percent difference in stress is decisive here only because the everyday insult happens to land in the gap.

There is a second thickness effect, in time rather than stress, and it is larger. Thermal diffusivity is conductivity divided by density times specific heat, which for soda lime glass is 1.0 divided by 2500 times 840, or 4.76e-7 square metres per second. The time for a wall to even out is roughly its thickness squared divided by that. For 2 millimetres: 4e-6 over 4.76e-7, which is 8.4 seconds. For 6 millimetres: 3.6e-5 over 4.76e-7, which is 75.6 seconds. Three times the thickness, nine times the duration, because it goes as the square. So the thick glass is not only more stressed, it is held there nine times longer, and a crack only needs one unlucky flaw during that window.

One refinement the usual story misses entirely, and it is where tumblers actually break. A heavy tumbler is rarely uniformly thick. It is typically a thin wall on a deliberately thick base, and those two sections develop completely different gradients on completely different timescales, so they try to expand by different amounts while rigidly joined. That junction is also a geometric stress concentrator. If you have broken tumblers this way, look at where they went: it is usually a ring at the base rather than a vertical split up the wall, and that is differential expansion between two thicknesses rather than a gradient through one.

Tempering is a different trick entirely

Which brings us to the third bar on the chart, because it reaches the same place by a route that has nothing to do with expansion coefficients.

Tempered glass is heated close to softening and then cooled rapidly at the surfaces. The surfaces solidify first and the interior contracts afterwards, which leaves the finished part with its surfaces locked in compression and its core in a balancing tension. Fully tempered glass carries at least 69 megapascals of surface compression and commonly around 100.

Now the consequence. Any tensile stress applied from outside has to cancel that compression before it can put the surface into net tension and open a flaw. So the effective tensile capacity is the inherent 40 megapascals plus roughly 100 megapascals of prestress, or about 140 in total. Divide that by soda lime's 0.808 megapascals per kelvin and you get 173 kelvin. Tempered ordinary glass therefore tolerates a slightly larger temperature step than borosilicate does, using the same high expansion glass, purely because it was cooled cleverly.

And the trade-off, which is the thing to understand before trusting it. The core is in tension, at roughly half the surface compression, so there is a large amount of stored elastic energy in the part. While the compressive skin is intact, nothing can reach that core and the glass is tough. Once a crack does penetrate past the compression layer into the tension zone, the stored energy drives it to branch, and branch again, until the whole piece dices into small fragments in a few milliseconds. That is why tempered glass does not crack and carry on, and why a chipped edge or a deep scratch is a much more serious matter on tempered glass than on annealed glass: the chip has already breached the only part that was protecting you.

Our three borosilicate products, and what each publishes

The worklist for this article warned that only one of our products names borosilicate and that the page might have to be abandoned for lack of support. Three do, and the term appears 95 times across 25 published pages, so the opposite problem turned out to be true: the word is everywhere and the reason for it is nowhere.

A set of FlexiTuff borosilicate glass food storage containers publishes five containers and five lids, 35.2 ounces each, 8.2 by 6.3 by 2.7 inches, airtight leak-proof lids, BPA-free borosilicate, and microwave and freezer safe. In its pros it claims the set "Handles temperature changes", which on everything above is the correct claim for the material and is worth 151 kelvin rather than 49.5.

And then the thing worth flagging, which its own record flags first: the cons say "Only microwave and freezer safety stated". So here is a borosilicate container whose single distinguishing material property is thermal shock tolerance, and whose listing declines to state an oven rating. Microwave and freezer are the two undemanding cases. The oven is the one where borosilicate earns its place over soda lime, and it is the one the listing is silent on. A reader could reasonably conclude from the material that it is ovenproof and should not, because the listing does not say so. Our FlexiTuff borosilicate glass food storage containers review has the detail.

A Zulay Kitchen glass electric kettle publishes 1.7 litres, 1500 watts, a "Borosilicate glass and stainless steel" body, auto shut-off, boil-dry protection, a cool-touch handle and a 360 degree cordless base. A kettle is the clearest case for this material in a kitchen, because it does the 80 kelvin step twice a day for years, and fatigue from repeated sub-critical thermal cycling is real for glass even when no single cycle breaks it. The record's cons are sensible and physical: glass "shows limescale, wants regular descaling" and is "less knock-proof than metal or plastic". The second is the one that connects to this page, because mechanical surface damage is what sets glass's tensile strength, so a knocked and scratched kettle has a lower thermal shock threshold than a new one even though its material has not changed. See our Zulay Kitchen glass electric kettle review.

A corrupted record in our own database

An OVENTE 1.5L glass electric kettle belongs here for the material, publishing a borosilicate body, 1100 watts, 1.5 litres, auto shut-off, boil-dry protection and a 360 degree rotary base. Those figures are read from its summary and pros, deliberately, because its specifications field is unusable and that is worth reporting.

All four entries in that product's specification list carry the label "Feature" rather than a field name, and all four values are Amazon bullet headings truncated mid-word: "A BLEND OF STYLE and DURABILITY , The OVENTE KG83 Kett", "ENJOY HOT WATER ANYTIME , This attractive glass elec", "CONVENIENCE AT ITS FINEST , Enjoy peace of mind with", and "US-BASED CUSTOMER SERVICE LETS YOU BUY WITH CONFIDEN". Its pros field is also stored as one unbroken run of text where every other product on the site uses one claim per line.

That is the same corruption pattern already identified in our four dishwasher product records, and it had not been noticed outside that category. It means this product publishes no capacity, no wattage and no material in any structured field, and a reader who reads only the specification table learns nothing about it at all. The figures are recoverable from the prose, which is what this page did, and the record needs rebuilding. Our OVENTE 1.5L glass electric kettle review is where that shows.

The tempered case we carry, and a spec that undercuts itself

A Leopard tempered glass cutting board is our one tempered product, at 16 by 12 inches on non-slip rubber feet, and it demonstrates both halves of the tempering trade.

Its specification field says, simply, "Heat: Heat resistant". Its cons field says "Heat resistant, not proof against thermal shock". The cons are right and the specification is the kind of claim this whole page exists to take apart, because heat resistance and thermal shock resistance are different properties and tempered glass has a great deal of one and a moderate amount of the other. On the arithmetic above its tempering is worth about 173 kelvin of step, which is genuinely good, and the reason the cons are still correct is the failure mode: when tempered glass does go, it goes completely and into fragments, so "resistant until it is not" is a fair description and a bare "heat resistant" is not.

Worth adding one thing the listing does not. A glass board accumulates exactly the kind of surface damage that sets tensile strength, because a knife is a scratch-generating instrument and the record itself notes the board is "loud under a blade" and that "hard glass can dull knives faster". Chips at the edge are the specific hazard for a tempered part, since the compressive skin is the only thing holding the core's stored energy in check. Our Leopard tempered glass cutting board review covers the rest.

Our own page, and the sentence that gives the right material the wrong reason

Our guide to choosing glass food storage containers for meal prep reaches the right recommendation and supports it with the wrong facts. Under a heading telling you to favour borosilicate, it says: "Borosilicate resists staining and odors that plastic absorbs over time. It is also BPA-free, making it a cleaner choice for storing food day after day."

Every clause there is true and not one of them is about borosilicate. All glass is chemically inert, so all glass resists staining and odours. All glass is BPA-free, because BPA is a monomer used in certain plastics and glass contains no polymer at all. Soda lime glass does every single thing that sentence credits to borosilicate, and does it just as well. So the guide recommends the right material while listing only properties that would equally justify the ordinary one, and omits the single property that is genuinely unique to it, which is the factor of 3.06 in thermal stress.

The same guide does get one thing right in a way worth keeping: it says to "Look for microwave and freezer safe glass that handles temperature changes. That lets you freeze a batch and reheat it in the same dish." That is the correct use case and it is exactly the one borosilicate is for, since going from a freezer to a microwave is a large and fast temperature step. The guide just needs to connect the two.

The pages of ours that already had it right

Two deserve the credit, and both are roundups rather than guides. Our kettles roundup says of a glass model that "borosilicate glass resists thermal shock and lets you see the water and its level at a glance", and lists "resists thermal shock" among its pros. Correct, correctly attributed to the material, and the right property named. Our pizza stones coverage is better still, distinguishing materials by this exact property: "cordierite tolerates the fierce temperatures of a grill and shrugs off thermal shock, while clay and ceramic suit steady home-oven baking". That is thermal shock resistance used as a genuine selection criterion between two materials, which is precisely how it should be used.

And our general storage comparison, glass versus plastic food storage containers, gets the practical instruction right: "The catch is thermal shock: moving glass straight from the freezer into a hot oven can crack it, so let frozen glass thaw a bit or follow the maker's guidance." Right advice, right hazard, and the only thing missing is why thawing works, which is that it reduces the temperature difference rather than the temperature. Thermal shock does not care how hot your oven is. It cares about the gap.

The honest limit

Not one product we carry publishes an expansion coefficient, a maximum temperature differential, a tempering stress, or a wall thickness. Three name borosilicate and none says what that buys. Twenty one published pages of ours use the phrase thermal shock and none explains it. So the figures here are handbook properties for the two glass families and a stated assumption about working tensile strength, and the one genuinely uncertain input is that strength, which depends on how scratched your particular glass is and could reasonably be anywhere from 30 to 50 megapascals. That moves the soda lime threshold between about 37 and 62 kelvin, which is a wide band and is itself part of the answer to why this feels unpredictable.

The free techniques beat the purchase comfortably, because every one of them works on the temperature difference rather than on the glass. Warm the glass first: half an inch of hot tap water, swirled and poured away, brings the whole wall up together and takes perhaps 40 kelvin straight off the step. Pour slowly and down the side rather than straight onto the base, which spreads the arriving heat over more area and avoids the thick base entirely. Let boiling water stand for a minute, since dropping from 100 to 90 degrees removes an eighth of the difference for no effort. Never put a glass straight from the fridge under a hot tap, which is the same insult with the sign reversed and the cold glass starting further away. And put a metal spoon in a glass before pouring, which is the old trick and does genuinely work, not by magic but by conducting heat away from the first contact point and spreading it, which lowers the local gradient while the pour is in progress.

Then the one that matters most and is free in a different sense: stop running thin glassware through a dishwasher loaded against other glass. Tensile strength here is a surface flaw property, not a material property, so every chip and scratch is a permanent reduction in the temperature step that glass will ever survive again. A scratched soda lime tumbler is not the same object it was when you bought it, and no amount of careful pouring gives that back.