Skip to content

Buckingham Pi Theorem

Intuition

Dimensional analysis by guess-a-power-law works, but how do you know how many independent dimensionless combinations exist, and whether you found them all? Buckingham's theorem answers with linear algebra: count variables, subtract the number of independent dimensions, and that's exactly how many dimensionless \(\pi\)-groups describe the problem.

Theorem

If a physical relationship involves \(n\) variables built from \(k\) independent fundamental dimensions, then it can be rewritten as a relationship among

\[\boxed{\,n - k\ \text{dimensionless groups}\,}\]
\[f(q_1, \dots, q_n) = 0 \quad\Longleftrightarrow\quad F(\pi_1, \dots, \pi_{n-k}) = 0\]

where each \(\pi_i = q_1^{a_1} q_2^{a_2}\cdots q_n^{a_n}\) is dimensionless. The exponent vectors span the null space of the dimensional matrix — hence the count.

Procedure

  1. List variables (this is the physics step — everything after is mechanical)
  2. Write dimensions of each in \([M], [L], [T], \dots\)
  3. Build the dimensional matrix (rows = dimensions, columns = variables)
  4. Find the null space — each basis vector is one \(\pi\)-group
  5. Interpret the groups (they usually have names!)

Practical shortcut: pick \(k\) "repeating variables" that span the dimensions (commonly a length, a velocity, a density), then combine each remaining variable with them.

Worked example: wind turbine force

Force \(F\) on a turbine tower given wind speed \(v\), air density \(\rho\), viscosity \(\mu\), rotor diameter \(d\), spin rate \(N\). Here \(n = 6\), \(k = 3\) ⇒ 3 groups. Choosing \(d, v, \rho\) as repeaters:

\[0 = \phi\left(\underbrace{\frac{F}{\rho v^2 d^2}}_{\pi_1:\ \text{force coefficient}},\; \underbrace{\frac{\rho v d}{\mu}}_{\pi_2:\ \text{Reynolds number}},\; \underbrace{\frac{dN}{v}}_{\pi_3:\ \text{tip-speed ratio}}\right)\]

Six dimensional variables collapse to a surface in 3-D dimensionless space — this is why one wind-tunnel campaign can characterize every geometrically similar turbine.

Worked example: pipe flow rate

\(Q = \phi(D, L, \Delta P, \mu)\): \(n = 5\), \(k = 3\) ⇒ 2 groups:

\[\pi_1 = \frac{Q\mu}{\Delta P D^3} \quad (\text{flow characteristic}), \qquad \pi_2 = \frac{D}{L} \quad (\text{aspect ratio})\]

Adding the physical insight that \(Q\) depends on the gradient \(\Delta P/L\) forces the combination \(\pi_1 \propto \pi_2\), i.e.

\[Q \sim \frac{\Delta P\, D^4}{\mu L}\]

— the Hagen–Poiseuille law up to the exact prefactor \(\pi/128\) (for diameter) that only the full solution supplies.

Common mistakes

  • Miscounting \(k\). It is the number of independent dimensions actually present — check the rank of the dimensional matrix, not just "3".
  • Choosing repeating variables that cannot span the dimensions (e.g. two lengths) — the algebra will fail.
  • Treating the \(\pi\)-groups as unique. Any product of powers of groups is another valid set; pick the conventional, physically named ones.

Knowledge graph position

Prerequisites: Dimensional analysis, linear algebra (null space). Leads to: Reynolds number, dynamic similarity, turbomachinery scaling.

Quiz

Q1 (computational). Drag \(F\) on a sphere depends on \(\rho, v, D, \mu\). How many \(\pi\)-groups, and what are they?

Answer

\(n = 5\), \(k = 3\) ⇒ 2 groups: drag coefficient \(C_D = \frac{F}{\rho v^2 D^2}\) and Reynolds number \(Re = \frac{\rho v D}{\mu}\). Hence \(C_D = f(Re)\) — the entire sphere-drag literature is one curve.

Q2 (conceptual). For the simple pendulum with \(T, L, m, g\) (\(n=4\), \(k=3\)) the theorem promises one group. What does the single group tell you?

Answer

\(\pi = T^2 g/L\) = const ⇒ \(T \propto \sqrt{L/g}\), and since no dimensionless combination can include \(m\) alone, the period cannot depend on mass.

Q3 (multiple choice). Two flows are dynamically similar when:

  • (a) they have the same fluid (b) all their \(\pi\)-groups match
  • (c) their velocities match (d) their forces match
Answer

(b). Geometric similarity plus equality of every relevant dimensionless group (in practice often just \(Re\), plus Mach/Froude/Weber as relevant).