Torque is a proxy, and not a very good one
What holds a bolted joint together is clamp load: the tension in the bolt, pulling the two parts against each other hard enough that they never move relative to one another. That tension is what you want to control. What a torque wrench measures is the twisting effort needed to turn the nut, and the overwhelming majority of that effort goes into overcoming friction in the threads and under the bearing face. Only something like ten to fifteen percent of the input turns into bolt stretch.
All of the messiness follows from that ratio. Because friction dominates, anything that changes friction changes the tension you get from a given torque, and by a lot. The whole of that variability is bundled into one empirical number, the nut factor K, in the relationship torque equals K times nominal diameter times clamp load. K is not a physical constant. It is a fudge that happens to work, and published values scatter by a wide margin even for nominally identical hardware.
The nut factor, and the mistake it causes
| Thread condition | Typical K | Effect vs dry |
|---|---|---|
| Stainless on stainless, dry | 0.28 | +40% torque for the same load |
| Zinc plated, dry | 0.22 | +10% |
| Plain steel, as received | 0.20 | baseline |
| Black oxide | 0.18 | -10% |
| Machine oil, light grease | 0.15 | -25% |
| Wax, moly, anti-seize | 0.12 | -40% |
Read that table as the answer to a question people ask constantly: what happens if I put anti-seize on a bolt and use the dry torque spec from the chart. The answer is that the clamp load comes out about forty percent higher than intended. A fastener aimed at 75 percent of proof lands past 100 percent, which is past yield, and it either takes a permanent stretch or snaps. Lubricated hardware needs a lubricated torque figure, and a spec that does not state the thread condition is incomplete.
Even with the condition pinned down, torque control is only good to roughly plus or minus twenty five to thirty percent on the resulting preload. That is why the target sits at 75 percent of proof rather than at 100: the margin absorbs the scatter. Methods that measure the bolt rather than the friction, such as turn-of-nut, ultrasonic length measurement, load-indicating washers or hydraulic tensioning, are what get used when the preload genuinely has to be known.
Stress area, and why it is not the bolt diameter
The bolt does not fail across its nominal diameter, because the threads cut into it. Nor does it fail at the minor diameter, because the helix means the failure surface is somewhere in between. The convention is the tensile stress area, computed for unified threads as 0.7854 times the square of the major diameter minus 0.9743 divided by threads per inch, and for metric threads as 0.7854 times the square of the nominal diameter minus 0.9382 times the pitch.
Working an example through: a 1/2-13 bolt gives 0.7854 x (0.5 - 0.0749) squared, which is 0.1419 square inches. At Grade 5 with a proof strength of 85,000 psi the proof load is 12,061 pounds, three quarters of that is 9,046 pounds of clamp force, and at K of 0.20 the torque is 0.20 x 0.5 x 9046, or 905 pound-inches, which is 75 pound-feet. That agrees with the published dry figure for a 1/2-13 Grade 5 fastener, which is where every torque chart you have ever seen comes from.
Fine threads have a larger stress area than coarse threads of the same nominal size, because there is less material cut away, so a 1/2-20 takes more torque and carries more load than a 1/2-13 of the same grade. That is one of the few places where the finer thread is genuinely stronger rather than just feeling more precise.
Where a chart is the wrong tool
A generic torque figure assumes the bolt is the weakest thing in the joint. Very often it is not. Threads tapped into aluminum strip long before a Grade 8 bolt yields. A gasketed flange has a specified sequence and a specified torque that has nothing to do with the bolt's proof load and everything to do with crushing the gasket evenly. Anything with a manufacturer specification, from cylinder heads to wheel nuts to structural connections, has that specification for reasons that are not visible from the fastener alone, and the specification wins over any chart including this one.
Reused fasteners are their own problem. A bolt that has been to 75 percent of proof once is usually fine; one that has been overtightened, heated, corroded or has a damaged thread is not, and none of that is visible at the wrench. Torque-to-yield bolts, which are common in engines, are designed to be taken past yield exactly once and are not reusable at all.
Related shop arithmetic
The hole comes before the fastener: the tap drill size calculator sizes it for the engagement you want, and thread engagement percentage interacts directly with how much of this clamp load a tapped hole can actually take. For the parts being bolted together, the metal weight calculator covers stock weight and the thermal expansion calculator covers what happens to a bolted joint that gets hot.
Questions people ask
What is the torque for a 1/2-13 Grade 5 bolt?
About 75 lb-ft dry, as received. That comes from a tensile stress area of 0.1419 square inches, a Grade 5 proof strength of 85 ksi giving a proof load of 12,061 pounds, a target of 75 percent of that at 9,046 pounds of clamp force, and a nut factor of 0.20. Lubricate the same bolt and it drops to about 57 lb-ft for the same clamp load, and applying 75 lb-ft to a greased bolt overloads it. Grade 8 in the same size is around 106 lb-ft dry because the proof strength is 120 ksi instead of 85.
Should I torque a bolt dry or lubricated?
Whichever the specification says, and if there is no specification, be consistent and use the matching K value. Lubricated joints are more repeatable because the friction scatter is smaller, which is a real advantage, and they are also easier to take apart later. The trap is applying a dry chart figure to a lubricated fastener, which overshoots the preload by around a third to a half. If you lubricate, reduce the torque accordingly. Note also that lubricant on the underside of the head matters as much as lubricant in the threads, because roughly half the friction is at the bearing face.
What percentage of proof load should I aim for?
75 percent is the common target for a reusable joint tightened with a torque wrench, and this page defaults to it. The reasoning is that torque control scatters the actual preload by roughly a quarter either way, so a 75 percent target keeps the high end of the scatter below yield. Structural steelwork tightened by turn-of-nut goes higher, often to 90 percent or beyond, because the method measures the bolt's rotation rather than the friction and is far more accurate. Joints that see fatigue loading want high preload, because a well-preloaded bolt sees almost none of the fluctuating external load; the bolts that break in service are usually the loose ones.
Why does the same bolt have different torque figures in different charts?
Because the charts assume different nut factors, different preload targets, or both, and many do not state either. A chart built on 65 percent of proof with K at 0.20 and one built on 75 percent with K at 0.18 will differ by about a quarter on the same fastener, and neither is wrong. Some automotive charts also quote figures for a specific plating or a specific thread locker. When two sources disagree, the useful move is to work out the clamp load each implies and see which one matches your joint, rather than averaging the two torques.
Does this apply to bolts threaded into aluminum?
The bolt side of the calculation does, and it is usually not the limiting factor. Threads tapped into aluminum, magnesium or plastic strip at loads well below what a steel bolt can carry, so the joint is governed by the tapped material and by how many threads are engaged rather than by the bolt grade. The general rule for a tapped hole in a softer material is that engagement of two to three times the bolt diameter is needed to approach the bolt's own strength, and the correct answer for a specific case comes from the shear area of the internal thread, not from a torque chart. Where the housing has a published spec, use it.