Dimensional Analysis: Why n − r Groups Exist

The Dimensional Analysis panel of the Science Lab hub has five tabs: check an equation for homogeneity, identify the quantity behind a unit, compute with units in a plain expression, reduce a variable list with the Buckingham Π theorem, and audit the formula library the rest of the product uses. It works on seven base dimensions — mass, length, time, current, temperature, amount and luminous intensity — and keeps the exponents as exact rationals.

Updated 2026-10-07Reviewed by CalcX Editorial Team5 sources4 min read

How to work a tab

Check an equation.Write both sides around = and give each symbol a quantity; the panel prints the dimension of each side and any missing factor.
Identify a unit.Type a unit such as N or a quantity name to see its dimension vector and every named quantity that shares it.
Compute with units.Type ordinary arithmetic with unit suffixes: 5 m/s * 3 s, 5 km + 300 m, or convert with to.
Reduce a model.In the Buckingham tab, list each variable's name and quantity; the count of independent dimensionless groups appears above the Π table.
Audit the library.The audit tab runs the same check over the product's own formula list and flags every equation whose sides do not match.

Worked example

Wave drag on a hull: variable list F (force), ρ (density), v (velocity), L (length), g (acceleration) gives n = 5 variables and rank r = 3, so the theorem predicts 2 independent groups:

Π group	Expression	Exponents (F, ρ, v, L, g)	Dimension
Π₁	F/(ρ·v²·L²)	(1, −1, −2, −2, 0)	1
Π₂	F·g²/(ρ·v⁶)	(1, −1, −6, 0, 2)	1

The other presets show the count at work: pipe flow with ρ, v, D and μ has n = 4, r = 3, one group — Π₁ = ρ·v·D/μ, the Reynolds number. Convection with h, L, k, ρ, v, μ and c_p has n = 7, r = 4 and 3 groups — h·L/k, L·ρ·v/μ and h/(ρ·v·c_p). Unit arithmetic is checked by the same machinery: 5 m/s * 3 s reads 15 m, 5 km + 300 m reads 5.3 km, 100 degC to degF reads 212 degF, and (1/2) * 1200 kg * (30 m/s)^2 to kJ reads 540 kJ.

Homogeneity is a filter, not a proof

Every physical variable is a point in a seven-dimensional exponent space, and a valid equation must add terms that sit at the same point — F = m*a works because both sides are M·L·T⁻², while F = m*v leaves a stray T and the check names the missing factor. The Buckingham theorem goes further: form the matrix of variable dimensions — one column per variable, one row per base dimension; its rank r counts the dimension directions the variables actually span, and n − r independent exponent vectors survive in its null space. Each survivor is a dimensionless product, and any physical law must relate those products alone — which is why an experiment on wave drag reduces to a curve of one Π against the other, whatever the units. The groups are a basis, not a canon: wave drag's Π₂ can be traded for the Froude number v²/(gL) with the same span.

Limits and what it does not do

Dimensions, not physics.: Homogeneity cannot tell you whether the constant is 2 or ½π, whether a sign is wrong, or whether the model is causal; it only rejects what cannot be true under any scaling.
The library audit is only as good as its mappings.: A formula with no unit metadata is reported as unmapped rather than failed, so a clean audit is not proof of correctness.
What it does not do.: It does not convert a full dataset or fit a dimensioned function — for numerical work use the Error Propagation panel, for chemistry quantities Molar Mass, and for the units themselves the converter.

The trap: one dimension, several quantities

Dimension vectors collide. Energy, work, heat and torque all reduce to M·L²·T⁻², so typing that vector returns a list, not a name — the panel says so instead of picking one. The second trap is that n − r is not n minus the number of base dimensions you happened to use: it is the rank of the actual variable matrix. Drop a variable that carried a direction and the count rises; the convection preset's seven variables span only four directions, so three groups remain even though four base dimensions appear in the table.

Where it is used

Catching algebra slips

Check every derived line the moment it is written. A missing factor of a length shows up as a dimension mismatch before any number is substituted, and the suggested missing factor often points straight at the term that was dropped.

Planning a scale experiment

Before building a model, list the variables and read n − r. Two groups mean the experiment is a family of curves in one parameter; three mean a surface. The Π table is the design sheet for the test matrix.

References

Wikipedia, Buckingham π theorem — n variables, rank r, n − r groups. en.wikipedia.org （访问日期：2026-10-07）
Wikipedia, Dimensional analysis — homogeneity and scale modelling. en.wikipedia.org （访问日期：2026-10-07）
NIST, Special Publication 811 — SI units for dimension vectors. nist.gov （访问日期：2026-10-07）
BIPM, The International System of Units (SI Brochure) — the seven base quantities. bipm.org （访问日期：2026-10-07）
Wikipedia, Reynolds number — a worked dimensionless group. en.wikipedia.org （访问日期：2026-10-07）