Why a Lubricated Bolt Snaps at the Right Torque

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

No, and the question is worth answering carefully because the premise is exactly backwards in the direction that breaks things. A lubricated bolt reaches a higher clamping force at the same torque, not a lower one, and on a small fastener the difference is enough to take it past the load it is designed to survive. The same wrench, at the same setting, on the same bolt: dry it lands at 87 percent of the bolt's proof load, and oiled it exceeds the bolt's ultimate tensile strength and snaps. Neither reading tells you which happened, because the wrench was accurate to within 3 percent the whole time.

A bolt is a spring, and tightness is not a thing it has

Start with what a bolt does. It is not a pin that holds two parts together by being in the way. It is a spring that you deliberately stretch, and the tension stored in that stretch squeezes the joint faces against each other. That tension is called preload, and preload is what resists the joint coming loose, what carries vibration without the fastener walking out, and what fails when a joint fails.

Put a number on the stretch, because it is smaller than almost anyone guesses. An M8 bolt has a tensile stress area of 36.6 square millimetres. Take 30 millimetres of grip length and steel at 200 gigapascals, and the bolt's spring rate is 200e9 times 36.6e-6 divided by 0.030, which is 244 kilonewtons per millimetre. To reach a sensible 15 kilonewtons of preload you therefore have to stretch it by 15,000 divided by 244,000, which is 0.061 millimetres. Sixty one microns. A human hair is around seventy.

That is the entire job. Everything a torque wrench exists to do is to stretch a piece of steel by rather less than the width of a hair, and torque is not a measurement of that stretch. It is a measurement of how hard you had to twist to produce it, which is a different quantity with a large and unpublished term in the middle.

Where the torque actually goes

The long form relationship between the torque you apply and the preload you get is this: the torque equals the preload multiplied by the sum of three lengths. The first is the thread pitch divided by two pi, which is the part that genuinely converts rotation into stretch. The second is the thread friction radius divided by the cosine of the thread flank angle. The third is the friction radius under the bolt head where it bears on the joint.

Work it for a standard M8 by 1.25 at a friction coefficient of 0.15, which is a reasonable figure for clean dry steel. The pitch is 1.25 millimetres, so the first term is 1.25 divided by 6.283, which is 0.199 millimetres. The pitch diameter is 8 minus 0.6495 times 1.25, which is 7.188 millimetres, so the thread friction radius is 3.594 millimetres, and with a 30 degree flank half angle the second term is 0.15 times 3.594 divided by 0.866, which is 0.6225 millimetres. A standard hex head is about 13 millimetres across the bearing face against an 8.4 millimetre hole, giving a mean bearing radius near 5.35 millimetres, so the third term is 0.15 times 5.35, which is 0.8025 millimetres.

Add them: 0.199 plus 0.6225 plus 0.8025 is 1.624 millimetres. Now read off the proportions. The term that stretches the bolt is 0.199 of 1.624, which is 12.3 percent. Thread friction is 38.3 percent. Head friction is 49.4 percent. So roughly seven eighths of the effort you put into a torque wrench is converted directly into heat in two sliding contacts, and about one eighth does the thing you wanted.

One correction to how this is usually quoted, including in my own notes for this article. The familiar rule of thumb is "about 50 percent in the threads, 40 percent under the head, 10 percent stretching". The 10 percent is right, near enough. The other two are the wrong way round for a standard hex head, because the head bears on a larger radius than the threads act at, and friction torque scales with radius. Which of the two dominates genuinely depends on head style, a flange head pushing the head term higher and a small socket cap head pulling it lower, so the ordering is not universal. The part that does not move much is the useful fraction, and it is always close to a tenth.

The nut factor, and the number that snaps the bolt

Divide that 1.624 millimetre sum by the 8 millimetre nominal diameter and you get 0.203. That dimensionless number is the nut factor, written K, and it collapses the whole mess into the short form every fastener table uses: torque equals K times diameter times preload. Rearranged, preload equals torque divided by K times diameter. K is the only term on the right that nobody publishes and that you cannot see.

Now change one thing. Put light oil on the thread and under the head, taking the friction coefficient from 0.15 to 0.08. The pitch term does not move, because geometry does not care about oil. The thread term becomes 0.08 times 3.594 divided by 0.866, which is 0.332. The head term becomes 0.08 times 5.35, which is 0.428. The sum is 0.199 plus 0.332 plus 0.428, which is 0.959 millimetres, so K is 0.959 divided by 8, which is 0.120.

Take 30 newton metres, the top of the range of the one wrench we carry that publishes a range at all, and run it both ways on that M8.

Dry, at K equals 0.203: preload is 30 divided by 0.203 times 0.008, which is 18,470 newtons. Call it 18.5 kilonewtons.

Oiled, at K equals 0.120: preload is 30 divided by 0.120 times 0.008, which is 31,250 newtons. Call it 31.3 kilonewtons.

Those have to be set against what the bolt can take. An M8 in property class 8.8 has a proof strength of 580 megapascals, so over 36.6 square millimetres its proof load is 21.2 kilonewtons, and a tensile strength of 800 megapascals, so its ultimate load is 29.3 kilonewtons. The dry figure, 18.5, is 87 percent of proof, which is roughly where a designer would want it. The oiled figure, 31.3, is 147 percent of proof and 107 percent of ultimate. The bolt does not hold tighter. It fails.

What this does to a published accuracy figure

This is where the whole category stops making sense, and our own catalogue supplies the numbers.

A DSBYKJ 17 piece torque wrench set is the only product we carry that publishes a torque range rather than a single figure, and it is the one worth arguing from: 5 to 30 newton metres on a quarter inch drive, plus or minus 3 percent accuracy, a 72 tooth reversible ratchet and a stated 10,000 cycle test. Take that accuracy claim seriously. At 30 newton metres, plus or minus 3 percent is plus or minus 0.9 newton metres, which propagates into plus or minus 3 percent of preload, so about plus or minus 550 newtons on the dry figure above.

Set that against the friction swing. The instrument is specified to plus or minus 550 newtons. The lubrication condition, which nobody specifies anywhere, moved the answer by 12,800 newtons. The uncontrolled variable is 23 times larger than the instrument's entire error budget. That is not a criticism of the wrench, which appears to be a careful piece of equipment and is doing precisely what it claims. It is the observation that the care has been spent on the term that was already small.

Its own record is honest about the limits that matter for using it, noting the 5 to 30 newton metre range is for bikes, cameras and electronics rather than car lug nuts, that it must be stored wound down to protect calibration, and that it will need re-checking over time like any torque wrench. All three are true and all three are about the instrument. Our DSBYKJ torque wrench set review has the detail. What no listing in the category publishes, this one included, is the condition under which its reading means anything.

Our own guide, and the sentence to correct

We have one dedicated guide to this tool and it is good on everything it covers. What is a torque wrench, how to use one and why the range matters is right that range selection is the beginner's trap, right that working at the very bottom or top of a range is inaccurate, and right about storing a click wrench wound down. Hold on to that last point, because it is a clue the guide does not follow: it explains that winding down "relaxes the internal spring and preserves accuracy". The guide knows there is a spring in the wrench. It does not notice there is a bigger one in the bolt.

The sentence to correct is this one: "Tightening becomes a measurement, not a guess." It is the most quotable line in the guide and it is half true in a way that matters. Tightening does become a measurement. The trouble is that it becomes a measurement of torque, and torque is not the quantity you care about, and the conversion between the two runs through a coefficient that is not on the box, not in the manual and not knowable by looking. Twelve percent measurement, 88 percent guess, with the guess dressed as a number.

The guide also says "Every important bolt has a correct torque set by the manufacturer." A torque figure without a stated condition is not a specification, and this is why real torque tables carry a condition in the same row: dry, oiled, lightly oiled, with thread locker, with anti-seize. The same joint can legitimately carry two torque figures that differ by nearly a factor of two, and both are correct, because they are specifying the same preload through two different nut factors. If you have ever wondered why a workshop manual bothers to say "clean and dry", that is the whole reason, and it is not housekeeping advice.

Three tools that change the sum, and one that does not

Three of the accessories we carry act on this arithmetic, and it is worth being precise about which term each one touches, because the market is not.

A flexible 360 socket extension bar set is the useful case to be careful about. A straight extension between the wrench and the socket changes nothing at all: torque is transmitted along a shaft without loss, so a six inch extension delivers what you set. Its record is clear-eyed about what it actually is, calling it "a reach tool with sensible torque limits" and "not for very high-torque or seized bolts", which is the correct framing. But a flexible or universal joint is a different animal. Once the drive runs at an angle, part of the input goes into the joint itself rather than into the fastener, and the loss grows with the angle, so the figure at the wrench is no longer the figure at the bolt. The honest rule is that you can tighten through a bend or you can tighten to a number, and not both. Details in our flexible 360 socket extension bar set review.

An Olutacot 4 piece adjustable tap wrench set belongs in this argument for a reason the listing does not give. It cuts the thread, and thread quality is one of the inputs to the friction coefficient that decides everything above. A clean, correctly cut, correctly aligned thread has a repeatable friction condition. A torn, galled or cross-threaded one does not, and no amount of instrument accuracy recovers from it. The record is unusually candid here, noting that "thread quality still depends on user technique" and that "best results need careful alignment checks". Both are true and both matter more than they sound, because a bad thread does not just feel rough, it moves the nut factor. See our Olutacot adjustable tap wrench set review.

And the case where none of this applies, which is worth saying so the page does not overreach. A linseyCove multifunctional sink and faucet wrench is for the mounting nuts behind a sink, and those are not preloaded fasteners in the sense used here. They compress a gasket to make a seal, so the target is a sealing pressure rather than a stretched bolt, and the correct technique is genuinely feel plus a quarter turn after the gasket seats. There is no torque figure to miss. The record notes it is "a specialist plumbing tool, not general DIY" and that "badly seized nuts can still be a fight", which is exactly right, and a seized nut is the one place where a wrench's job is to exceed a friction torque rather than to meter one. Our linseyCove sink and faucet wrench review covers it.

Stainless is the worst case, and galling is why

One material deserves its own paragraph because the arithmetic above goes further wrong on it than on anything else, and because our own patio furniture listings quietly depend on it.

Austenitic stainless steels such as 304 run a much higher friction coefficient against themselves than plain steel does, commonly 0.3 or worse, and they do something plain steel mostly does not. Both surfaces carry a thin passive oxide film and are relatively soft and ductile underneath. Slide them together under pressure, the film breaks through at the high spots, clean metal meets clean metal, and the two surfaces cold weld. The weld then tears, roughens the contact, and the friction climbs as you turn. That is galling, and it is a runaway: the more you tighten, the worse the condition that is resisting you gets.

Put it in the same terms as the rest of this page. At a friction coefficient of 0.35 the nut factor for that M8 comes out near 0.42, so 30 newton metres gives 30 divided by 0.42 times 0.008, which is 8,900 newtons. Barely half the dry steel figure, from the same setting on the same wrench. And that is the optimistic reading, because it assumes a constant coefficient, whereas a galling thread's coefficient is rising while you work. The click you are waiting for can arrive because the thread is seizing rather than because the joint is tight, and those two feel identical through a wrench.

Which is why the standard practice on stainless is anti-seize or a nickel or copper based lubricant, and why that in turn drags you straight back to the first half of this page: the lubricant that stops the galling also drops the nut factor, so the torque figure has to come down with it. "Galling" appears nowhere across our 2,527 published pages, and neither does "anti-seize". That matters more than a vocabulary gap, because several of our outdoor listings specify 304 stainless hardware as a selling point, which it is for weather resistance and which is simultaneously the worst case for tightening by feel. Both things are true of the same screw and no listing mentions the second.

Problems in our own records

Two things turned up that are worth reporting rather than burying.

The first is a gap rather than an error, and it is total. Across every published product we carry, not one fastener or fastening tool publishes a friction condition, a nut factor, or a preload figure. Torque appears 77 times across our published pages and every instance is a capacity rating or a product name. So the single number that converts what a torque wrench reads into what a joint actually gets is absent from the entire catalogue, and as far as we can tell from the listings, from the entire category. That is not a failing of any one product.

The second is a definitional muddle in our own vise and clamp coverage, which is where the phrase "clamping force" lives on this site. Nine roundups rate products in clamping force, up to 660 pounds on one portable bench vise and up to 7 kilonewtons on a drill press vise, and best-portable-bench-vises defines it as how hard "the jaws squeeze". That definition is correct for a vise and it is a different quantity from the clamping force in this article, which is the tension in a stretched bolt. They share a name and nothing else. Worth flagging because the two are easy to conflate, and a reader who has met the phrase on our vise pages will arrive here expecting it to mean jaw pressure.

The honest limit, and the free technique that beats the purchase

The specification that would settle all of this is preload, measured directly. It is measurable, by bolt elongation with a micrometer, by ultrasonic length measurement, or by an instrumented load washer, and all three are out of reach for the jobs these tools are bought for. So you are left controlling preload through a proxy that is seven eighths noise, and the realistic ambition is not accuracy.

What you can have is consistency, which is worth more than it sounds. Most joints with several fasteners fail because one is loose or one is overloaded rather than because the average is wrong, and an even set of eight mediocre preloads beats a scattered set containing one perfect one. So: take every thread to the same condition, which in practice means clean and dry unless a manual says otherwise, and never lubricate a thread whose torque figure was written for a dry one. Tighten in the sequence the manual gives, and tighten in two or three passes rather than straight to the final figure, because the first fasteners you tighten are relaxed by the ones you tighten afterwards as the joint settles.

And then the technique that costs nothing and beats the instrument on the quantity that matters. Snug every fastener until the joint faces are genuinely in contact with no gap, which is the only part that needs feel. Then mark the socket or the nut with a pen and turn a measured angle. Because the thread is a screw, the angle directly sets the advance of the nut, and the advance sets the stretch. The 61 microns this page opened with is about 18 degrees of rotation on a 1.25 millimetre pitch, and because the joint members compress a little as well as the bolt stretching, call it somewhere above 20 degrees in practice. On a hex head, one flat is 60 degrees, so the whole operation is under half a flat past snug.

The point of doing it that way is not that the angle is easy to get right. It is that friction has dropped out of the measurement entirely. A protractor on a socket is reading the stretch, which is what holds the joint. A torque wrench is reading the stretch plus seven times as much friction, and then reporting the total to within 3 percent.