Drag and lift
A body sitting in a moving fluid feels a force, and asking for its size in newtons is the wrong first question. The right one is what fraction of it comes from friction along the surface and what fraction from a pressure imbalance built by the flow separating behind it — because those two fractions call for entirely different fixes.
Every force a moving fluid exerts on an immersed body comes from exactly two sources: the pressure acting normal to the surface, and the shear stress acting tangential to it. Add up the components of both, in the direction of the oncoming flow, and the total is drag. Add up the components perpendicular to it and the total is lift. The two forces are the same integral, read in two different directions, and the whole of this article is about how the balance between pressure and friction, and where the boundary layer separates, decides which one you get and how large it turns out to be.
Pressure drag and friction drag
Formally, drag is the sum of two surface integrals, one over the pressure and one over the wall shear stress, each projected onto the flow direction:
Neither term is optional and neither is usually negligible on its own, but which one dominates depends entirely on the body's shape, and the two extremes are worth having a picture of. A long, thin, well-aligned aerofoil keeps its boundary layer attached along almost its whole length. The pressure recovers on the back nearly as much as it rose on the front, so the two pressure integrals very nearly cancel, and what is left is overwhelmingly the friction acting along a large wetted area.
A bluff body — a sphere, a flat plate held square to the flow, the blunt front of a truck — cannot do that. Its surface curves away from the flow too sharply for the boundary layer to follow, so it separates well before reaching the back, leaving a wide, low-pressure, unsteady wake there. The high pressure built up at the front, where the flow is being brought nearly to rest, is never matched by a corresponding recovery at the back, and that imbalance is pressure drag, sometimes called form drag. For a body like this it swamps the friction term, which is why sanding a cannonball smooth barely changes how hard it is to throw against the wind.
The distinction earns its keep in design. Reducing friction drag means treating the surface itself — polishing it, keeping the boundary layer laminar for as long as possible, since a laminar layer's skin friction is lower than a turbulent one's at the same Reynolds number. Reducing pressure drag means reshaping the body: moving the point of maximum thickness rearward, tapering the tail gradually enough that the flow can follow it without separating. Applying a friction fix to a pressure-dominated problem, or the reverse, wastes effort on the smaller of the two numbers.
On a streamlined body the flow stays attached from nose to tail.
The drag coefficient
A raw drag force in newtons tells you almost nothing on its own, because it mixes together the body's size, the fluid's density, and how fast the two are moving relative to each other. Divide it by a reference pressure built from exactly those quantities, and by the area the flow is presented with, and what remains is a shape-dependent number with the size and the speed divided back out:
The reference pressure \(\tfrac12\rho V^2\) is the dynamic pressure met already in Bernoulli’s equation, and \(A\) is, by the usual convention for a drag-type coefficient, the frontal area — the silhouette the body presents facing into the flow, not its total wetted surface. Getting that area right matters more than it looks: quoting the same \(C_D\) against two different area conventions is a standing source of confusion between data sheets.
Once nondimensionalised this way, shapes become directly comparable regardless of their size or the speed they were tested at, provided the Reynolds number is in a broadly similar range. A flat plate held square to the flow has a \(C_D\) of roughly 1.2; a sphere, roughly 0.47; a well-streamlined strut or fuselage shape, as low as 0.04. That last number is not a typo relative to the first two — it is the entire argument for why the next section exists.
Every body's drag depends on its size, its speed, and the fluid around it.
Streamlining and the drag crisis
Streamlining a bluff body — tapering its rear into a long, gentle tail rather than leaving it blunt — works by giving the boundary layer an adverse pressure gradient gentle enough to survive without separating. Done well, it can shrink a wide, turbulent wake down to almost nothing, at the cost of a larger wetted surface and therefore more friction drag. Up to a point that trade is overwhelmingly favourable, since pressure drag on a bluff body is so much larger than the friction it is trading against; taken too far, the added surface starts to cost more in friction than it saves in pressure, and the total drag turns back upward.
A stranger version of the same idea happens without changing the shape at all. Plot the drag coefficient of a smooth sphere against Reynolds number and, across a wide sub-critical range, it sits close to a steady \(C_D \approx 0.47\). Then, within a narrow band around \(Re \approx 3\times10^5\), it falls off a cliff to roughly \(0.1\), a drop of a factor of four or five over a small change in speed. This is the drag crisis, and the mechanism is exactly the transition met in boundary layers: below the critical Reynolds number the layer stays laminar all the way round the front of the sphere and separates early, leaving a wide wake. Above it, the layer trips into turbulence before it separates, and a turbulent layer's vigorous internal mixing lets it cling to the surface further round the back, pushing the separation point aft and shrinking the wake substantially — enough to more than pay for the modest rise in skin friction that comes with turbulence.
This is also the honest explanation for the dimples on a golf ball. A smooth ball hit at typical speeds sits just below the critical Reynolds number, in the expensive, wide-wake regime. Small dimples trip the boundary layer into turbulence deliberately and early, buying the low-drag, narrow-wake state at a speed where a smooth ball could not reach it on its own. The rougher surface increases friction drag slightly and cuts pressure drag by far more, and the net result travels noticeably further.
Below a critical speed the layer stays laminar all the way to separation.
Lift
An aerofoil at a small positive angle of attack, or simply with enough camber, forces the air passing over its upper surface along a more curved, longer path than the air underneath in the same time. It is worth being careful with that last clause: the popular explanation that the two parcels of air must arrive at the trailing edge together, and so the upper one must move faster to cover its longer path, is not actually true and is not needed either. What is true, and is enough on its own, is simply that the flow over the top is measurably faster than the flow underneath. By Bernoulli’s equation, faster means lower pressure, so the upper surface runs at a pressure below the lower one:
Integrate that pressure difference over the whole surface, projected perpendicular to the oncoming flow, and what results is lift. In practice the majority of it comes from suction on the upper surface rather than from extra push on the lower one, which is why a wing's upper contour is the part an aerodynamicist worries about first.
The pressure picture is correct and it is also incomplete, because it does not say why the flow over the top should be faster in the first place. The rigorous statement is built on circulation: an aerofoil with camber or angle of attack sustains a net circulation \(\Gamma\) around its section, a value fixed not by guesswork but by the Kutta condition — the physical requirement that the flow leave a sharp trailing edge smoothly, rather than careering around it at infinite speed the way an inviscid solution without circulation would demand. Once \(\Gamma\) is fixed, the lift per unit span follows directly from the Kutta–Joukowski theorem:
The pressure argument and the circulation argument are not competing explanations; they are the same flow field described at two different levels, and the circulation reading is the one that survives contact with a proper derivation.
Camber and angle of attack force the air over the top along a tighter path.
The lift coefficient and stall
Lift nondimensionalises exactly as drag did, against the same dynamic pressure and, for a wing, the planform area rather than a frontal one:
For a thin aerofoil at modest angles, thin-aerofoil theory gives a strikingly simple result: \(C_L\) rises almost exactly linearly with angle of attack \(\alpha\), at a slope of close to \(2\pi\) per radian,
and that straight-line relationship is what every early exercise in lift is built on. It cannot continue forever. As \(\alpha\) increases, the upper surface is asked to support an ever-steeper adverse pressure gradient right behind its leading edge, and the same separation mechanism from boundary layers is fighting to break the flow loose. Somewhere around \(\alpha_{\text{stall}} \approx 15\text{-}18^\circ\) for a typical section, it wins.
The upper-surface flow separates suddenly and extensively, the neat pressure difference the linear theory relied on collapses with it, and \(C_L\) drops sharply from its peak value, typically somewhere near \(1.5\) for a simple unflapped section, at exactly the angle where a pilot or a designer would otherwise have wanted the most lift available. Drag rises sharply at the same angle, for the same reason: the separated wake that used to cost only a little pressure drag on a well-behaved wing now costs a great deal. Stall is not a separate phenomenon from everything else in this article. It is the same separation, arriving at the worst possible moment.