The fourth power in a pipe

Push fluid through a pipe and the rate depends on the pipe’s radius. Everyone expects that. What almost nobody expects is how steeply — the flow goes as the fourth power, so a pipe half the width carries a sixteenth as much.

The exponent is the whole content of this page, and it is worth seeing where each of its four powers comes from.

The balance that sets the shape #

Consider a cylinder of fluid of radius $s$, coaxial with a pipe of radius $R$ and of length $L$, in steady flow. It is not accelerating, so the forces on it balance.

Pushing it along is the pressure difference acting on its ends: $\Delta p \cdot \pi s^2$. Holding it back is the viscous stress on its curved surface, which by the definition of viscosity is $\mu\,\mathrm{d}u/\mathrm{d}s$ acting over an area $2\pi s L$. Setting them equal:

$$\Delta p \, \pi s^2 = -\mu \frac{\mathrm{d}u}{\mathrm{d}s} 2\pi s L, \qquad \frac{\mathrm{d}u}{\mathrm{d}s} = -\frac{\Delta p}{2\mu L} s.$$

The velocity gradient is proportional to the distance from the axis. Integrate, and apply the no-slip condition that $u = 0$ at $s = R$:

$$u(s) = \frac{\Delta p}{4\mu L}\left(R^2 - s^2\right).$$

A parabola. The fastest fluid is on the axis, the fluid at the wall is stationary, and everything in between varies smoothly. Nothing was assumed about the shape; it came out of a force balance and a boundary condition.

Where the four powers come from #

Now integrate the parabola over the cross-section to get the volume flow:

$$Q = \int_0^R u(s)\, 2\pi s \,\mathrm{d}s = \frac{\pi \Delta p R^4}{8 \mu L}.$$

That is the Hagen–Poiseuille law, and the $R^4$ is what everything else on this page is about. It arrives as a product of two separate factors of $R^2$, and separating them makes the exponent memorable rather than surprising:

Two powers from the area. A pipe of radius $R$ has cross-section $\pi R^2$, so twice the radius offers four times as much room.

Two powers from the speed. The peak velocity is $\Delta p R^2 / 4\mu L$ — proportional to $R^2$ as well. A wider pipe is not merely roomier; the fluid in it moves faster at every corresponding position, because the slowest fluid (at the wall) is further away from the middle, so the gradients required are gentler.

The second factor is the one people miss, and it is the one that turns an unremarkable dependence into a dramatic one.

The half that is exact #

The mean velocity over the cross-section is exactly half the peak, and the figure computes it by quadrature rather than quoting it.

That is a fact about integrating a paraboloid: the volume under a paraboloid of revolution is half the volume of the cylinder that contains it. It has nothing to do with fluids, viscosity or pressure, and it is the same geometric statement that makes the average of $R^2 - s^2$ over a disc equal to $R^2/2$.

It matters practically because a flow meter measuring the peak velocity on the axis — which is what many ultrasonic and Doppler meters do — must halve it to get a flow rate, and only in this regime. In a disorderly flow the profile is much flatter and the ratio is closer to 0.8, so applying the wrong factor is a twenty per cent error in a quantity somebody is billing for.

What the exponent does to the world #

Blood. Vessel resistance goes as $R^{-4}$, so the body can redistribute blood dramatically by changing radii slightly. A twenty per cent constriction multiplies resistance by 2.4; a twenty per cent dilation cuts it to 0.48. That sensitivity is why smooth muscle in arteriole walls is the principal control mechanism for blood distribution, and why the same exponent makes arterial narrowing so disproportionately serious.

Needles. Hypodermic gauge is a radius, and the force required to inject at a given rate goes as $R^{-4}$. Moving from a 21-gauge needle to a 25-gauge — a radius ratio of about 1.8 — multiplies the required force by roughly eleven, which is why fine needles are used for thin fluids and thick ones for viscous suspensions.

Water mains. Doubling a pipe’s diameter multiplies capacity by sixteen at the same pressure drop, so a modest increase in bore is often far cheaper than a pumping station. The same exponent is why partial blockage by scale is so costly: a pipe narrowed ten per cent by deposits has lost a third of its capacity.

Porous media. A soil or a filter is a bundle of channels of varying width, and because each contributes as the fourth power of its own radius, the flow is dominated by the widest few. A material’s permeability is therefore set by its largest channels rather than by its average pore size — which is why a small number of cracks can destroy the sealing performance of an otherwise fine-grained barrier.

What distinguishes a fourth power is not that it is large at any particular radius but how fast it amplifies a ratio. A factor of two in the radius is a factor of sixteen in the flow, and a factor of ten is ten thousand — so halving an artery’s diameter cuts its flow to a sixteenth, and a boiler tube scaled up by a third carries three times the water. The exponent is what makes small changes in bore into large changes in everything downstream.

Reading the viscosity off the flow #

Every quantity in the law except $\mu$ is straightforward to measure, so the relation is routinely run backwards and used as a viscometer. Drive a known pressure across a known tube, catch the fluid for a known time, weigh it, and $\mu$ follows.

Capillary viscometry has been the standard method for a long time and it is accurate, but the $R^4$ cuts both ways: an error in the radius is amplified fourfold in the answer. A one per cent error in bore is a four per cent error in viscosity, so the tubes are calibrated against a fluid of known viscosity rather than measured geometrically. What the instrument really provides is a ratio to a standard, and the fourth power cancels out of the ratio.

That is a general lesson about steep power laws used as instruments. High sensitivity to the quantity being measured is the same thing as high sensitivity to everything else, and a design that cannot separate the two is a design that amplifies its own errors. The usual escape is a comparison rather than an absolute reading — which is the same strategy as weighing a body in and out of water.

The resistance analogy, and where it breaks #

$Q = \Delta p \cdot (\pi R^4 / 8\mu L)$ has the same form as a current proportional to a voltage, so a pipe can be given a resistance $8\mu L/\pi R^4$, and networks of pipes can be analysed exactly like networks of resistors — in series the resistances add, in parallel the conductances do.

The analogy is genuinely useful and it has two limits worth naming. First, it holds only while the flow stays orderly; beyond that the relation between $\Delta p$ and $Q$ is no longer linear and no fixed resistance describes it. Second, it assumes the flow is fully developed — that the parabola has had length to establish itself. Near an entrance the profile is still flat and the resistance is higher; the entrance length is a few tens of radii in slow flow, which for a short wide pipe can be the whole thing.

There is a third limit that is specific to biology and is large: blood is not Newtonian, and in vessels comparable with a cell’s size it is not even a continuum. Red cells migrate away from the walls, leaving a thin plasma layer that lubricates the flow, so apparent viscosity falls as vessel diameter falls below about 300 microns. The effect reverses in the finest capillaries, where cells must deform to pass at all. Poiseuille himself was a physician measuring blood pressure, and the law he is named for describes his subject only approximately.

The pressure that drives it #

Nothing on this page says where $\Delta p$ comes from, and it is worth connecting to the rest of the ladder. In a horizontal pipe it is supplied by a pump. In a vertical one, gravity contributes $\rho g h$ by the ordinary hydrostatic law, which is why a tall header tank can drive a flow with no pump at all. In a syringe it is the plunger, and the force required is the pressure times the plunger’s area, which is why a hydraulic system’s geometry and this law interact in a design.

The power dissipated is $Q \Delta p$, and it all becomes heat — the same dissipation the previous rung identified, now integrated over a pipe. For a domestic water supply it is negligible. For a long crude-oil pipeline it is the dominant operating cost, and it is why such lines are heated: warming the oil lowers $\mu$, and the pumping power falls in proportion — an application of the steep temperature dependence a liquid’s viscosity has rather than of anything on this page.

An oscillating system trades energy back and forth and returns it; a viscous flow does not. The work done against the pressure drop in a pipe goes entirely into heat and none of it is recoverable — which is why a pipe has a pressure drop rather than a pressure difference that could be run backwards, and why pumping costs money continuously rather than once.

The same law in a slot, and why the exponent moves #

It is worth doing one other geometry, because it shows which part of the fourth power belongs to the physics and which to the circle.

For flow between two parallel plates a distance $h$ apart, the same balance gives a parabolic profile across the gap and a flow per unit width proportional to $h^3$. Three powers, not four: two from the profile as before, and only one from the cross-section, because widening the gap does not widen the channel in the other direction.

So the exponent is not a universal constant of viscous flow. It is two plus the number of directions in which the passage gets bigger, and a round pipe happens to grow in two. That reframing makes the result portable: a square duct gives four, an annulus of fixed circumference and growing gap gives three, and a fracture in rock gives three — which is why the permeability of a fractured rock is dominated by aperture in a way that a granular rock’s is not.

It also explains why lubrication works. A bearing’s film is a slot with a very small $h$, so the $h^3$ makes the resistance to squeezing out enormous, and a thin film can carry a large load for a long time. The same steep exponent that makes a narrowed artery dangerous is what makes an oil film load-bearing.

How a quantity scales with size is a question about how many directions grow with it, and that is the geometric half of the whole argument. A pipe gains two powers from its cross-section and two more from the velocity profile being pinned at the wall; a slot of the same depth gains one and two, and comes out as a cube. Same physics, different count of directions, different exponent.

The exponent that designs a circulation #

A resistance going as $R^{-4}$ makes wide pipes cheap to pump through, and a body cannot simply make every vessel wide: blood costs something to make and to carry, and a wider vessel holds more of it. Balancing those two is a one-line optimisation with a memorable answer.

The pumping power in a vessel is $Q^2$ times its resistance, so it falls as $R^{-4}$. The cost of maintaining the blood inside it rises as its volume, so it grows as $R^{2}$. Add the two, differentiate, and set the derivative to zero: the optimum has

$$Q \propto R^3.$$

Flow proportional to the cube of the radius. And since flow is conserved where a vessel divides, the radii at a bifurcation must satisfy $R_0^3 = R_1^3 + R_2^3$ — Murray’s law, from 1926, and it is obeyed to good accuracy by arteries, by the airways of a lung, by the xylem of a tree and by the tracheal tubes of an insect.

The corollary is the part that explains how a body could possibly implement it. Put $Q \propto R^3$ into the expression for the shear stress the flow exerts on the vessel wall, $4\mu Q/\pi R^3$, and the radius cancels: an optimally proportioned network has the same wall shear stress everywhere in it. So a vessel does not need to know the global optimisation. It needs only to sense the shear its own lining is feeling and to widen or narrow until that shear reaches a set point, which is precisely what endothelial cells do.

A global design problem has been reduced to a local feedback, and the reduction is available only because the resistance goes as the fourth power and the stress as the third.

The spreading that gets worse when diffusion gets better #

The parabola has a second consequence that is easy to overlook and matters wherever anything is carried along a pipe rather than merely pushed through it.

Inject a short plug of dye into a laminar flow. The fluid on the axis moves at twice the mean and the fluid at the wall does not move at all, so the plug is drawn out into a long spike — a paraboloid whose tip runs ahead while its skirt stays behind. After a modest distance the dye occupies a length far greater than it started with, and a chemist trying to keep a sample together is in trouble.

What actually happens is stranger, and better. Molecules also diffuse across the tube, and a molecule that spends some time on a fast streamline and some on a slow one travels at something near the average of the two. Given long enough for a molecule to cross the tube many times, every molecule has sampled every speed, and the plug stops stretching linearly and begins to spread diffusively about the mean position — with an effective diffusion coefficient

$$D_{\text{eff}} = \frac{R^2 \bar u^2}{48 D}.$$

The molecular diffusivity is in the denominator. Making the solute diffuse faster across the tube makes the plug spread more slowly along it, because the averaging is more complete. Taylor found this in 1953 and it remains one of the more counterintuitive results in transport.

The size of it is what makes it practical. For water in a millimetre-bore tube at a millimetre a second, the effective coefficient is some twenty thousand times the molecular one. Every chromatography column, every flow-injection analyser and every microfluidic device is designed around it — narrow bores, because $D_{\text{eff}}$ goes as $R^2$, and the same fourth-power law then says what that costs in pressure.

Where the model stops #

The flow must be orderly. Above a threshold set by a ratio of inertia to viscosity, the smooth parabola breaks down, the profile flattens, and the resistance rises sharply — flow then grows roughly as $\sqrt{\Delta p}$ rather than in proportion to it. Predicting where that happens, and describing what follows, belongs to the collection that owns flow regimes; what belongs here is that the law has a boundary and does not announce it.

The fluid is Newtonian. Blood, paint, polymer solutions and slurries are not, and for a shear-thinning fluid the profile is blunter than a parabola and the exponent differs from four. That is the next rung.

The flow is steady. Pulsatile flow — an artery, a reciprocating pump — has a profile that changes through the cycle and does not reduce to this one. The relevant comparison is between the pulse period and the time momentum takes to diffuse across the pipe, which is the penetration depth from the previous rung.

The pipe is straight, round, rigid and smooth. A bend adds resistance; a non-circular section changes the constant; an elastic tube widens under pressure, which makes $R$ depend on $\Delta p$ and the law nonlinear in a new way. Arteries are elastic tubes, and that is a large part of why arterial flow is a subject of its own.

Slip is absent. The derivation used $u(R) = 0$. In a rarefied gas or on some engineered surfaces that fails, and the flow exceeds this prediction. The gas case has a criterion attached: slip matters once the mean free path becomes comparable with the bore, which for air at atmospheric pressure means channels below a few microns.

The fluid is a continuum. Below that same scale the fluid stops being describable by a viscosity at all, and a flow becomes a problem about individual molecules crossing a channel.

The history, and two independent discoveries #

Gotthilf Hagen, a German hydraulic engineer, published the relation in 1839 from measurements on water in brass tubes. Jean Léonard Marie Poiseuille, a French physician interested in blood circulation, published it independently in 1840 from measurements on much finer tubes. Neither derived it; both measured it, and both found the fourth power empirically by varying the radius and watching the flow change far more than expected.

The derivation came later, once viscosity had been given the definition on the previous rung. That order is common here — Jurin’s capillary law had the same century-long gap between the measurement and the mechanism — and the same observation applies: the empirical law was right, and what the derivation added was reach rather than accuracy.

Poiseuille’s motive is worth keeping in view, because it is a good example of a measurement made for one purpose becoming a tool for another. He wanted to understand blood flow, obtained a law that describes blood flow only roughly, and gave the rest of physics an exact result about something else.

There is a sharper version of that worth stating. The law is exact for the idealisation and approximate for the case that motivated it, and the gap between the two is not a defect of the law but a statement about blood — that it is a suspension of deformable cells rather than a Newtonian liquid, and that vessels are elastic rather than rigid. Each departure names something real about the circulation. A model that fits everything explains nothing; one that fits a clean case exactly and a messy one approximately names exactly which features of the messy case are doing the work.

The ladder from here #

Later rungs on this anchor: pulsatile flow and the Womersley number. Flow between plates, in an annulus and in a non-circular duct, where the constant changes and the exponent does not. Entrance length and developing flow. The elastic tube, where radius depends on pressure. Lubrication theory, where a thin film of fluid carries an enormous load because the same viscous stresses act over a very small gap. And porous media, where a fourth power inside a distribution of channel sizes produces a permeability.