Friction
Almost every wrong friction answer comes from writing one equals sign too early. Static friction is not a force you can look up; it is a reaction, worth whatever equilibrium happens to need, and it becomes a known quantity only at the single instant the surfaces are about to slip.
Up to here every contacting surface has been smooth, and that was a deliberate simplification: a smooth surface can push only at right angles to itself, so it contributes exactly one unknown to the problem. Roughen it and the surface gains a second capability, a force along itself, and with it a second unknown. The equations do not get much harder. What gets harder is knowing what you are allowed to write down about that second force, and the answer is less than most people assume.
The other half of a contact force
Two solid bodies pressed together do not touch over the area they appear to share. They touch at a scatter of high points, and the true contact area is a small fraction of the apparent one. Each of those junctions pushes back in whatever direction its own local geometry dictates, and the interlocking and adhesion between them means those directions are not all normal to the nominal surface.
Add them up and you have a single resultant contact force \(\mathbf{R}\), generally tilted. In statics we never work with it directly. We resolve it into a component normal to the nominal surface and a component tangent to it:
\(N\) is the normal force, carrying the load pressing the bodies together. \(F\) is the friction force, and it acts in the plane of contact, opposing the direction in which the body would slide. This is dry, or Coulomb, friction: no lubricant film, no fluid. It is the case worth treating in statics because it is the one with a sharp, usable limit.
Two habits pay for themselves. The first is to draw \(N\) and \(F\) separately on the free-body diagram rather than drawing the tilted resultant, because the two components go into two different equilibrium equations. The second is to work out the direction of impending motion before drawing \(F\) at all, since friction opposes relative sliding, not the applied force. On a block being pushed up a slope, friction acts down the slope; on the same block sliding back down, it acts up. Same surfaces, opposite arrows.
Note also that \(F\) lies in the plane of contact, which need not be horizontal, and \(N\) is perpendicular to that plane, which need not be vertical. On an incline the weight is the only force that stays vertical. Resolving along and perpendicular to the surface, rather than along \(x\) and \(y\), is nearly always the shorter route.
Two rough surfaces touch at a scatter of high spots.
An inequality, not a formula
This is the section that decides whether the rest of the topic goes well. Experiment gives one result, and it is a limit rather than a value. The friction a pair of dry surfaces can supply before sliding begins is proportional to how hard they are pressed together:
The symbol that matters in that line is not \(\mu_s\). It is \(\le\). Static friction is a reaction like the force in a pin or the tension in a cord: it takes whatever value equilibrium demands of it, and only equilibrium tells you what that value is. Push a heavy crate with one finger and the friction on it is exactly the force of your finger — perhaps two newtons — even though \(\mu_s N\) for that crate might be four hundred. The surfaces are not holding anything in reserve for later. They are supplying two newtons because two newtons is what it takes.
So writing \(F = \mu_s N\) at the start of a problem is not a shortcut. It is an extra assumption, and if the body is merely sitting there in equilibrium the assumption is false and the answer will be wrong. There is exactly one circumstance in which the equality holds: when motion is impending — the instant at which the surfaces have run out of capacity and slipping is about to begin. Then, and only then,
Problems tell you which case you are in, and the wording is worth reading carefully. “Find the friction force on the block” is an equilibrium question: solve for \(F\) from \(\sum F = 0\) and then check it against \(\mu_s N\). “Find the smallest force that will move the block”, or “find the largest angle at which it will stay”, is an impending-motion question, and there the equality is not merely allowed but required — it is the extra equation that makes the problem solvable.
If a check comes back with the required \(F\) larger than \(\mu_s N\), the body is not in equilibrium at all. It is sliding, and the answer to report is that it moves, not some number. Once it is moving, kinetic friction takes over. That one is a genuine equality, \(F_k = \mu_k N\), with \(\mu_k\) a little smaller than \(\mu_s\) and roughly independent of speed, which is why a load that resists starting will then run away from you.
Two properties of \(\mu_s\) surprise people, and both follow from contact happening at high spots rather than over the whole face. It does not depend on the apparent contact area — a brick on its side and the same brick on its end need the same force to start — and it is a property of the pair of materials, not of either alone. Tabulated values are also crude, often quoted to one significant figure and varying by a factor of two with surface finish and contamination. Carrying a friction answer to five decimal places is arithmetic, not accuracy.
With nothing pushing, friction is zero. It is not waiting in reserve.
The angle and cone of friction
There is a second way to package the same physics, and for some problems it is far quicker. Instead of tracking \(N\) and \(F\) separately, keep the contact force whole and ask only how far it leans away from the normal. Call that angle \(\phi\). From the geometry of the resolution,
As the demand on the surface rises, \(F\) grows and \(R\) tilts further over. It can do so only until \(F\) reaches its limit, so \(\phi\) has a maximum of its own, the angle of friction:
In three dimensions the direction of impending sliding could be any direction in the plane of contact, so the limiting reaction sweeps a right circular cone about the normal, with half-angle \(\phi_s\). This is the cone of friction, and it converts the whole of the last section into a statement about direction: if the required contact force lies inside the cone, the body can be in equilibrium; on the cone, motion impends; outside it, no equilibrium is possible and the body slides.
The prettiest consequence is the angle of repose. Put a block on a plane and tilt it. Only two forces act — the weight and the single contact force — so equilibrium of two forces requires the contact force to be equal, opposite, and collinear with the weight: exactly vertical, whatever the slope is doing. The normal, meanwhile, tilts with the plane. The angle between them is therefore the angle of the plane itself, so on any incline
and the block slips at the tilt where \(\theta = \phi_s\). Tilting a plane until something slides measures \(\mu_s\) directly, with no force measurement and no knowledge of the weight, which is how the coefficient is often obtained in the first place. It is also why a heap of dry sand always settles to the same slope regardless of how much sand there is.
A block on a slope is held by one contact force and its own weight.
Tipping or sliding
Push on a tall wardrobe near the top and it does not slide; it goes over. Push near the floor and it slides. Something other than friction is deciding which happens, and a friction check alone will never find it. There are two independent ways for the body to stop being in equilibrium, and you have to test for both.
Sliding is a force question, and the last two sections answer it: work out the friction that equilibrium requires and compare it with \(\mu_s N\). For a block of weight \(W\) pushed horizontally by \(P\) on level ground, \(F = P\) and \(N = W\), so sliding impends when \(P = \mu_s W\). Note that the height at which you push does not appear.
Tipping is a moment question, and it turns on something usually left implicit: the normal force is the resultant of a pressure distributed over the whole base, and a resultant can act anywhere within the region carrying it. When \(P\) is applied at height \(h\), moment balance shifts that resultant forward, towards the edge the body would tip about, by
measured from the line of the weight. The pressure under the base can redistribute to produce that shift, but it cannot move the resultant beyond the base itself — the ground can push, and it cannot pull. So for a base of width \(b\) with the weight through the centre, the condition for not tipping is \(x \le b/2\), and tipping impends when
Now compare the two limits, \(\mu_s W\) and \(W b / 2h\), and the smaller one is what actually happens as \(P\) is raised from zero. Tall and narrow with a grippy floor tips; short and wide on ice slides. A useful way to hold it is that the two are equal when \(h = b/2\mu_s\) — push below that height and it slides, above it and it tips.
The commonest mistake is not the algebra but the assumption. Students set \(F = \mu_s N\), solve for \(P\), and report it as the answer without ever checking where the normal force ended up. If the required \(x\) exceeds \(b/2\), that sliding solution describes a body that tipped over before it ever got there. Always solve both, then choose.
A push above the base. Friction resists it; the normal force shifts.
Wedges and belts
Two applications are worth the space because they are where friction stops being a nuisance and becomes the mechanism. The first is the wedge — a block with a small taper, driven under a load to lift it or hold it. It is a machine that multiplies force, and friction is what decides whether it stays put afterwards.
A wedge problem is a set of free-body diagrams, one per body, with two or three contact surfaces each carrying its own \(N\) and its own \(F = \mu_s N\) at impending motion. The whole difficulty is bookkeeping: the direction of impending sliding at every surface has to be settled before any arrow is drawn, and for a wedge being driven in, every friction force opposes the motion of the surface it acts on, which is not always the same direction as the motion of the wedge. Get one arrow backwards and nothing downstream survives.
The design question is self-locking. If the taper angle is small enough — roughly, if it is less than twice the friction angle for the usual two-surface arrangement — then removing the driving force leaves the wedge in place, because the load alone cannot generate enough force along the faces to overcome friction. That is why a wedged door stop stays wedged and why a shallow screw thread does not unwind under load. Steepen the taper past that and the wedge shoots back out the moment you let go.
The second is a flexible band wrapped round a fixed drum: a rope on a bollard, a belt on a pulley, a brake band on a drum. Take a slice of the band subtending a small angle \(d\theta\). The tension either side pulls it against the drum with a normal force \(dN = T\, d\theta\), and at impending slipping that slice earns friction \(\mu\, dN\), which is precisely the amount by which the tension can change across it. So \(dT = \mu T\, d\theta\), and each slice multiplies rather than adds. Integrating over the total angle of wrap \(\beta\), in radians,
with \(T_2\) the larger tension, on the side towards which slipping impends. The exponential is what makes this worth knowing. With \(\mu = 0.3\), half a turn multiplies your grip by 2.6, a full turn by 6.6, and three turns by 285 — which is why a sailor can hold a boat against a bollard by hand, and why the drum’s radius appears nowhere in the formula at all. Only how far round the rope goes matters, not how big the thing is that it goes round.