Use dimensions to reject impossible formulas

Try

A formula adds a speed to an acceleration multiplied by time and calls the result a distance. Both terms on the right have speed units, so the result cannot be a distance. This defect can be found without collecting data or solving an equation.

Dimensions describe physical kinds such as length, time and mass. Units are chosen scales such as meters, seconds and grams. Changing units should change numerical representations consistently, not the physical prediction.

Predict

Suppose a pendulum’s characteristic time depends only on its length L and gravitational acceleration g. Predict a power-law form Lᵃgᵇ and determine the exponents. Can dimensional reasoning alone determine the familiar numerical factor or the dependence on a large release angle?

Experiment

Length has dimension length, while g has length divided by time squared. Therefore Lᵃgᵇ has dimensions length^(a+b) time^(−2b). Matching time gives a+b=0 and −2b=1, so a=1/2 and b=−1/2.

Proposed quantity	Dimension	Suitable as a time?
L/g	time squared	No
√(L/g)	time	Yes
√(g/L)	inverse time	No

The conclusion is a time scale proportional to √(L/g). The proportionality factor is dimensionless and can depend on dimensionless variables omitted from the initial list. Dimensional consistency is necessary, not a complete dynamical derivation.

Understand

Dimensionless groups can be found by linear algebra. Put the exponents of base dimensions for each variable into a matrix. A monomial is dimensionless when its exponent vector lies in that matrix’s nullspace. If there are n variables and the dimension matrix has rank r, the nullspace has dimension n−r. Under the usual positive-variable setting, independent null vectors generate independent dimensionless monomials. This is the algebraic mechanism behind Buckingham dimensional analysis.

For density ρ, speed U, length L and dynamic viscosity μ, the combination ρUL/μ is dimensionless. Its numerator and denominator both have mass divided by length-time units. A dimensionless relation may depend on this ratio, but dimensional analysis alone cannot determine that relation or guarantee that these are the only relevant physical variables.

Transcendental functions need dimensionless arguments. Write log(t/t₀), not an unexplained log of a dimensional time. A measured relation expressed in a fixed unit can still be a useful numerical convention, but the reference unit or scale must be explicit when converting or interpreting it.

Dimensional analysis also exposes parameter meanings. In x′=a−bx, a has stock-per-time units and b inverse-time units. Calling both “rates” without distinguishing their dimensions can hide an invalid parameter substitution. A fitted numerical coefficient is inseparable from its unit convention.

Apply

Audit every term of one equation from your investigation. Create a dimension table for states, inputs, parameters and outputs, then perform a unit conversion and verify that the reconstructed physical prediction is unchanged. Identify any dimensionless constants that still require theory or estimation; do not treat a dimensional check as a calibration.

Continue

Nondimensionalization uses these scales to rewrite the entire problem, including its initial and boundary data. It can reveal which combinations of parameters actually control the behavior.