Every derived quantity in mechanics can be expressed using just three base quantities — mass, length and time — written with the symbols $[M]$, $[L]$ and $[T]$. (For the full SI set we add $[A]$ for current, $[K]$ for temperature, $[\\text{mol}]$ and $[\\text{cd}]$.) The dimensions of a physical quantity are the powers to which these base quantities are raised to represent it. The expression that shows those powers is the dimensional formula.
For example, speed is length per time, so its dimensional formula is $[M^0L^1T^{-1}]$, usually shortened to $[LT^{-1}]$. Acceleration is speed per time, giving $[LT^{-2}]$. Force is mass times acceleration, so:
From force you can quickly build others: work or energy is force times distance, $[M^1L^2T^{-2}]$; power is energy per time, $[M^1L^2T^{-3}]$; pressure is force per area, $[M^1L^{-1}T^{-2}]$. A few quantities such as strain, refractive index and angle are pure ratios and are dimensionless, written $[M^0L^0T^0]$.
The single most useful rule is the principle of homogeneity: every term that is added, subtracted or equated in a physically correct equation must have the same dimensions. You cannot add a length to a time any more than you can add metres to seconds. This principle drives three powerful applications:
- Checking equations: if the dimensions on the two sides disagree, the equation is certainly wrong. (Agreement does not guarantee correctness, because dimensionless constants are invisible to the method.)
- Converting units between systems using the relation $n_1[M_1^a L_1^b T_1^c] = n_2[M_2^a L_2^b T_2^c]$.
- Deriving relations by assuming a quantity depends on others as a product of powers and matching dimensions on both sides.
The method has real limitations. It cannot find dimensionless constants (the $\\tfrac{1}{2}$ in $\\tfrac{1}{2}mv^2$ is undetectable), cannot handle equations with trigonometric, exponential or logarithmic functions (their arguments must themselves be dimensionless), and fails when a quantity depends on more than three other quantities since only three equations are available in mechanics.
Deeper Insight — homogeneity is a free error-detector: Treat dimensional analysis as a habit, not a chapter. Before trusting any derived formula in an exam, glance at its dimensions on both sides. If you wrongly write $v = u + \\tfrac{1}{2}at^2$, the term $at^2$ has dimensions $[LT^{-2}][T^2] = [L]$, which cannot be added to a velocity $[LT^{-1}]$ — the mistake reveals itself instantly. This single reflex catches a large fraction of careless algebra errors and costs you nothing.