🎯 Hook – measuring a nuclear bomb from a photograph
In 1950, physicist G. I. Taylor estimated the explosive yield of the first atomic bomb test using nothing but declassified photographs of the fireball at known times — no classified data at all. He reasoned that the fireball radius R could only depend on the energy released E, the air density ρ, and the time t, and dimensional analysis alone forced the relationship \(R \propto (Et^2/\rho)^{1/5}\). Checking which combination of quantities has the right dimensions can reveal a physical law before a single experiment is run.
📏 Dimensions and the principle of homogeneity
The dimension of a quantity shows how it depends on the base quantities, in square brackets: mass [M], length [L], time [T], current [I]. Working in SI base units is equivalent.
| Quantity | Defining equation | Dimensions | SI base units |
|---|---|---|---|
| Velocity | v = s/t | LT⁻¹ | m s⁻¹ |
| Force | F = ma | MLT⁻² | kg m s⁻² |
| Energy | E = Fs | ML²T⁻² | kg m² s⁻² |
| Pressure | p = F/A | ML⁻¹T⁻² | kg m⁻¹ s⁻² |
| Frequency, angular speed | f = 1/T | T⁻¹ | s⁻¹ |
So s = ut + ½at²: [L] = [LT⁻¹][T] + [LT⁻²][T²] = [L] + [L] ✔ — dimensionally consistent. Pure numbers and angles are dimensionless, and the argument of sin, cos, exp and log must always be dimensionless: in \(x=A\sin(\omega t)\), \(\omega t\) is dimensionless, so ω has dimensions T⁻¹; in \(N=N_0e^{-\lambda t}\), λ has dimensions T⁻¹.
🧮 Units of constants and deriving a relationship
Units of constants: rearrange the equation to make the constant the subject, then substitute dimensions. Example: \(G=Fr^2/(m_1m_2) \to \text{M}^{-1}\text{L}^3\text{T}^{-2}\).
Method of dimensions — to derive the form of a relationship:
- Guess which quantities the result depends on.
- Write \(y=kx_1^ax_2^bx_3^c\) (k dimensionless).
- Write the dimensions of both sides.
- Equate powers of M, L, T; solve for a, b, c.
🚧 Limitations
- Cannot find dimensionless constants — the 2π in \(T=2\pi\sqrt{l/g}\) never appears.
- Cannot distinguish quantities with the same dimensions (work and torque — both ML²T⁻²).
- Fails for sums of different terms or exponentials/logs of dimensional quantities.
- Needs a correct guess of the relevant variables; too many unknowns with too few base dimensions makes the solution non-unique.
- A dimensionally correct equation may still be wrong; a dimensionally incorrect one is definitely wrong.
📊 Graph mastery — testing a dimensionally derived power law
| Graph type | Axes (X, Y) | Shape | What to read |
|---|---|---|---|
| ln v vs ln T (testing v = k√(T/μ)) | ln T · ln v | Straight line | Gradient = predicted power (½ here); intercept gives k |
| v vs √T, v² vs T | √T or T · v or v² | Straight lines through the origin | Gradient related to k and μ |
✍️ Worked example 1 — the pendulum and a wave on a string
✍️ Worked example 2 — identifying a quantity from its units
🌍 Real-world: the Planck length
Dimensional analysis combines the fundamental constants \(\hbar\), G and c into the only quantity with dimensions of length: the Planck length, \(\sqrt{\hbar G/c^3}\approx1.6\times10^{-35}\) m — the scale at which quantum effects and gravity are expected to become comparably important.
🚀 Going further — Buckingham π theorem
With n variables and k independent dimensions, a physical law can always be written in terms of \(n-k\) dimensionless groups. Engineers use this to scale wind-tunnel models using the Reynolds number, a dimensionless combination of speed, size, density and viscosity.
Challenge: drag force F depends on ρ, v and A (3 variables, M L T = 3 dimensions) — how many independent dimensionless groups exist? \(n-k=3-3=0\): in this simple case F/(ρv²A) must itself be the single dimensionless constant k in F = kρv²A.
📋 Quick practice — answers in italics
(a) Write the dimensions of impulse. Impulse = Δp, so MLT⁻¹.
(b) Check \(v=\sqrt{2gh}\) for homogeneity. [LT⁻¹]² should equal [LT⁻²][L] = [L²T⁻²] ✔ — consistent.
(c) Find the SI base units of the Planck constant h from \(E=hf\). \(h=E/f\to(\text{kg m}^2\text{s}^{-2})/\text{s}^{-1}=\text{kg m}^2\text{s}^{-1}\).
(d) In \(x=A\cos(\omega t+\varphi)\), state the dimensions of ω and φ. Both must be dimensionless-when-multiplied-by-their-partner: ω has T⁻¹, φ is dimensionless (an angle).
⚠️ Trap corner – common mistakes
| Misconception | Correct view |
|---|---|
| Dimensions and units are different checks | [MLT⁻²] and kg m s⁻² carry the same information — use whichever notation the question asks for |
| Numbers and angles carry dimensions | Pure numbers, ratios and angles are always dimensionless — they drop out of a dimensional check entirely |
| Dimensionally correct means correct | A dimensionally correct equation can still have the wrong constant or even the wrong physics (E = mc² and E = 2mc² are both dimensionally fine) |
| Terms of different dimensions can be added if the numbers work out | Never add or equate terms of different dimensions — it is meaningless, whatever the numbers say (e.g. ½mv + mgh is invalid) |
| "Per" doesn't change the sign of a power | "Per" in words means a negative power: kg per m³ = kg m⁻³ |
| rad s⁻¹ has dimension T⁻² like acceleration | The radian is dimensionless, so rad s⁻¹ has dimensions T⁻¹, same as ordinary frequency |
Complete the statements:
✓ (b) A speed — specifically the speed of sound in the gas.
✓ (b) The dimensionless constant k (≈½C_d) cannot be found by this method.
✓ (c) At very low speeds, viscous (Stokes) drag dominates, which is proportional to v, not v² — a different physical regime the derivation didn't assume.
✓ (b) \(E=\tfrac12I\omega^2\) gives \(\text{ML}^2\text{T}^{-2}\)✔ — consistent with the translational analogy \(\tfrac12mv^2\), with I replacing m and ω replacing v, both squared.
✓ (c) Homogeneity only checks that dimensions balance; it cannot catch a wrong dimensionless constant or a wrong combination of otherwise correctly-dimensioned terms.
| T/N | v/m s⁻¹ |
|---|---|
| 1.00 | 14.3 |
| 2.00 | 19.8 |
| 4.00 | 28.0 |
| 6.00 | 35.0 |
| 8.00 | 40.3 |
✓ (b) ln T: 0, 0.693, 1.386, 1.792, 2.079. ln v: 2.660, 2.986, 3.332, 3.555, 3.696.
✓ (c) Gradient \(\approx(3.696-2.660)/(2.079-0)\approx\mathbf{0.50}\) — matches the predicted ½ well.
✓ (d) Intercept at ln T = 0 gives ln v = 2.660, so \(k\mu^{-1/2}=e^{2.660}=14.3\Rightarrow k=14.3\sqrt{0.005}\approx\mathbf{1.0}\).
✓ (e) \(v^2\) vs T gives a straight line through the origin with gradient \(k^2/\mu\).
✓ A true physical law describes a real relationship between quantities, so every term that is added or set equal must represent the same kind of physical quantity, hence the same dimensions.
✓ Cannot distinguish same-dimension quantities — work and torque are both ML²T⁻² but are physically different.
✓ Fails for sums of different terms or for exponentials/logs of dimensional quantities — it cannot analyse \(E_k+E_p\) term by term in the same way.
✓ Needs a correct guess of every relevant variable — an omitted or extra variable gives a wrong or non-unique answer.
✓ Measuring R and t from photographs at known times, and knowing ρ for air, allowed E to be estimated without any classified information.
✓ This required assuming no other variable (ambient pressure, radiation losses, geometry) significantly affected R — a simplification that happened to work well.
✓ Adding different powers of a dimensional quantity together is not dimensionally valid (x and x² would have different dimensions), so the argument must be a pure, dimensionless number for the series — and the function — to make sense.
✓ This shows dimensions describe only how a quantity scales with mass, length and time — not its full physical meaning; two quantities can share dimensions without being the same thing, so dimensional analysis alone can never fully identify a quantity.