Dimensional Analysis II: Deriving Relations & Limits

In 1947, Life magazine printed declassified photographs of the Trinity test — the world's first nuclear explosion — each frame stamped with a time in milliseconds and a distance scale. The bomb's energy yield was one of the most closely guarded secrets on Earth. The British physicist G. I. Taylor took a ruler to those public photographs and asked one innocent question: if the fireball radius $R$ can depend only on the energy released, the time elapsed, and the density of air, what form must the law $R(E, t, \rho)$ take? Dimensions alone forced the answer, and the answer forced out the secret: roughly $10^{14}$ joules — about 25 kilotons. No spy, no laboratory. Just the requirement that metres must be built out of metres. This article is Part II of dimensional analysis. In Part I you used dimensions as a detector, catching wrong equations by their mismatched units. Here you turn the same bookkeeping into a generator: assume a power-law form, balance the powers of $[M]$ , $[L]$ and $[T]$ , and out comes the formula for a pendulum's period, a wave's speed on a string, a satellite's orbital period — results you can derive years before you meet the full theory behind them. CBSE and ICSE ask this as a standard 3-mark derivation; JEE and NEET compress it into 60-second multiple-choice questions. The method also has famous blind spots — it cannot see the $2\pi$ , it chokes on sums and sines, and it fails when the variables outnumber the equations. Examiners test the failures as eagerly as the successes, so this article treats both as first-class citizens.

Key takeaways

To derive a relation, assume a power law $Q = k\,q_1^{a}q_2^{b}q_3^{c}$ , substitute dimensions, and equate the powers of $[M]$ , $[L]$ , $[T]$ on both sides — three simultaneous equations for the exponents.
The classic: $T = k\sqrt{L/g}$ for a simple pendulum. The mass exponent solves to zero — the method itself discovers that mass is irrelevant.
The same bookkeeping converts units between systems via base-unit ratios raised to the dimensional exponents: $1\ \mathrm{N} = 10^5\ \mathrm{dyne}$ in one line.
The method can never find dimensionless constants ( $k = 2\pi$ for the pendulum), cannot handle sums of terms, cannot see dimensionless quantities like angles, and fails when unknowns outnumber the three equations.
Taylor's blast-wave law $R = k\,(Et^2/\rho)^{1/5}$ is the method at full power — a state secret extracted from photographs.
Key idea: if a quantity can only be built from a given set of variables, demanding that both sides carry identical dimensions often pins down the entire law — everything except a pure number.

The core insight

Assume the unknown law is a product of powers, balance the dimensions $[M]$ , $[L]$ , $[T]$ on both sides, and solve for the exponents: dimensional consistency alone then dictates the form of the relation, leaving only a dimensionless constant that physics — not dimensions — must supply.

Where this lives today

Dimensional derivation never retired; it moved into the largest and most expensive physics. Wind tunnels exist because of it: a model aircraft at the right Reynolds number is, dimensionally, the full-size aircraft, so a metre of balsa predicts a hundred tonnes of aluminium. In biology, Kleiber's law — metabolic rate scaling as body mass to the $\tfrac34$ power from mice to whales — is a power law hunted down by exactly this exponent-balancing logic. In machine learning, the celebrated "scaling laws" linking a model's loss to parameter count and compute are fitted power laws whose exponents are debated with the seriousness Rayleigh reserved for sound. Working physicists reach for the method first because they know its boundaries: it brackets an answer before any simulation runs, flags impossible results afterwards, and occasionally replaces a classified database with one photograph and a square of graph paper.

The physical picture

Hang a bob from a string and set it swinging gently. What could its period $T$ depend on? The candidates: mass $m$ , length $L$ , gravity $g$ , amplitude. Now think like nature. The period is a time, so the formula must assemble pure seconds from those ingredients. Kilograms are a problem: neither $L$ nor $g$ contains any mass dimension, so if $m$ entered there would be nothing to cancel its kilograms — and a time cannot carry them. Mass must drop out. The amplitude is an angle, dimensionless, so the machine is silent about it. That leaves $L$ (metres) and $g$ (metres per second squared): the only way to build seconds is to divide them — $L/g$ is a time squared — and take a square root.

Figure

The exponent machine — dimensions force the pendulum formula

That is the whole method in one picture: list the ingredients, demand the dimensions balance, solve for the exponents. It feels like getting physics for free, and in a sense it is — you are cashing in the deepest structural fact about physical law, that equations must balance dimensionally no matter what units humans choose. The same picture scales up to Taylor's fireball: radius from energy, time and air density — three ingredients, three dimensions, one forced answer.

Figure

Taylor's fireball — one photograph plus dimensions equals a state secret

The recipe: one assumption, three equations

The engine behind every derivation here is a single assumed form: if $Q$ is suspected to depend on $q_1, q_2, q_3$ , write it as a product of unknown powers. Why a product? Because units combine multiplicatively — doubling a length doubles metres, never adds them — so any relation built purely from dimensional bookkeeping must be a product of powers. That is the method's engine and, as we will see, its cage.

Q = k\, q_1^{\,a}\, q_2^{\,b}\, q_3^{\,c} \qquad (1)

where: $k$ — a dimensionless pure number (invisible to the method, supplied later by experiment or full theory); $a, b, c$ — unknown exponents to be found; $q_1, q_2, q_3$ — the physical variables you judge relevant. Substituting each $q_i$ 's dimensional formula and matching the powers of $[M]$ , $[L]$ , $[T]$ against $Q$ yields up to three simultaneous equations — which is why the method works cleanly for three unknown exponents and fails for four. And the output is only as good as the ingredient list: right variables yield the law, a wrong list yields a dimensionally tidy falsehood. Board questions give the list ("assuming $T$ depends on $L$ and $g$ …"); in competitive exams choosing it is the question.

Derivation AThe pendulum period from dimensions alone

We want the period $T$ of a simple pendulum, assuming it can depend on the bob's mass $m$ , the string length $L$ , and the gravitational acceleration $g$ . The strategy is the recipe of equation (1): assume $T = k\,m^{c}L^{a}g^{b}$ , substitute dimensional formulae, and equate powers of each base dimension. Watch two things: how the mass exponent is forced to zero by pure bookkeeping, and how the final answer arrives with the constant $k$ still unknown.

Step 1. Assume the power law: $T = k\, m^{c} L^{a} g^{b}$ with $k$ dimensionless.

Step 2. Write the dimensional formula of every player: $[T] = [\mathrm{T}]$ , $[m] = [\mathrm{M}]$ , $[L] = [\mathrm{L}]$ , $[g] = [\mathrm{L\,T^{-2}}]$ .

Step 3. Substitute into the assumed form: $[\mathrm{T}^{1}] = [\mathrm{M}]^{c}\,[\mathrm{L}]^{a}\,[\mathrm{L\,T^{-2}}]^{b} = [\mathrm{M}^{c}\,\mathrm{L}^{a+b}\,\mathrm{T}^{-2b}]$ .

Step 4. Equate the powers of each dimension. For $[\mathrm{M}]$ : $0 = c$ . For $[\mathrm{T}]$ : $1 = -2b$ , so $b = -\tfrac12$ . For $[\mathrm{L}]$ : $0 = a + b$ , so $a = +\tfrac12$ . Why the mass vanished: $T$ contains no kilograms, and neither $L$ nor $g$ could supply or cancel any — so the only consistent mass exponent is zero. The method has independently discovered Galileo's observation that heavy and light bobs swing in step.

Step 5. Assemble the result: $T = k\, L^{1/2} g^{-1/2} = k\sqrt{L/g}$ . Full theory (simple harmonic motion, next year's work) supplies $k = 2\pi$ , giving $T = 2\pi\sqrt{L/g}$ . Why we cannot do better: $k$ is a pure number; multiply it by anything dimensionless and equation (1) still balances perfectly. Dimensions constrain structure, never pure numbers. *Verified: with $L = 1.0\ \mathrm{m}$ , $g = 9.8\ \mathrm{m/s^2}$ : $T = 2\pi\sqrt{1/9.8} = 2.01\ \mathrm{s}$ — the measured ≈2 s of a metre-long pendulum.*

Changing unit systems without conversion tables

The second exam-grade application looks different but runs on the same fuel. A physical quantity is a number times a unit, and the physical quantity itself does not care which unit you picked. Measure a force in newtons or in dyne — the force is the same, so the product of numerical value and unit size is invariant:

n_1 u_1 = n_2 u_2 \qquad (2)

where: $n_1, n_2$ — the numerical values in system 1 and system 2; $u_1, u_2$ — the sizes of the units. Read it physically: a bigger unit needs a smaller number ( $n \propto 1/u$ ) — 1 metre is the same length as 100 centimetres precisely because the centimetre is 100 times smaller. This inverse relation is the sanity check that catches nearly every conversion mistake. For a derived unit, the dimensional formula $[\mathrm{M}^{a}\mathrm{L}^{b}\mathrm{T}^{c}]$ tells you exactly how the unit is built from the base units, so the conversion factor is the product of the base-unit ratios raised to those powers:

n_2 = n_1\left[\frac{M_1}{M_2}\right]^{a}\left[\frac{L_1}{L_2}\right]^{b}\left[\frac{T_1}{T_2}\right]^{c} \qquad (3)

where: $M_1/M_2$ — the ratio of the old mass unit to the new one (similarly for length and time); $a, b, c$ — the exponents of the quantity's dimensional formula. The ratios point old-over-new: converting SI → CGS, $M_1/M_2 = 1\ \mathrm{kg}/1\ \mathrm{g} = 1000$ . If the answer moves the wrong way, test the formula on 1 m → 100 cm before hunting deeper bugs.

Go deeper · Derivation B · 1 minConverting one newton into dyne

Derivation BConverting one newton into dyne

We want the numerical value of 1 newton expressed in dyne, the CGS unit of force, using equation (3) rather than memorised factors. The route: recall force's dimensional formula $[\mathrm{M\,L\,T^{-2}}]$ , form the three base-unit ratios between SI and CGS, and raise each to the matching power. Watch the direction of each ratio — old system on top.

Step 1. Force has $a = 1$ , $b = 1$ , $c = -2$ in $[\mathrm{M}^{a}\mathrm{L}^{b}\mathrm{T}^{c}]$ , and $n_1 = 1$ in SI.

Step 2. Base-unit ratios, SI over CGS: $\dfrac{M_1}{M_2} = \dfrac{1\ \mathrm{kg}}{1\ \mathrm{g}} = 10^{3}$ , $\dfrac{L_1}{L_2} = \dfrac{1\ \mathrm{m}}{1\ \mathrm{cm}} = 10^{2}$ , $\dfrac{T_1}{T_2} = \dfrac{1\ \mathrm{s}}{1\ \mathrm{s}} = 1$ .

Step 3. Apply equation (3): $n_2 = 1 \times (10^{3})^{1} (10^{2})^{1} (1)^{-2} = 10^{5}$ . Why the time ratio was harmless: both systems share the second, so even the exponent $-2$ acts on a ratio of 1.

Step 4. State the result with its unit: $1\ \mathrm{N} = 10^{5}\ \mathrm{dyne}$ . Verified: unpacking directly, $1\ \mathrm{N} = 1\ \mathrm{kg\,m\,s^{-2}} = (1000\ \mathrm{g})(100\ \mathrm{cm})\,\mathrm{s^{-2}} = 10^{5}\ \mathrm{g\,cm\,s^{-2}} = 10^{5}\ \mathrm{dyne}$ . The same machinery gives $1\ \mathrm{J} = (10^{3})(10^{2})^{2} = 10^{7}\ \mathrm{erg}$ — recompute it as a drill.

The showpiece: reading a bomb from a photograph

Taylor's 1950 calculation is the same three-line algebra as the pendulum, wearing history's most dramatic clothes. A nuclear explosion dumps energy $E$ essentially at a point; at time $t$ the blast has "forgotten" every detail except $E$ and the density $\rho$ of the air it shoves aside — so the radius can depend on those three quantities alone.

Go deeper · Derivation C · 1 minThe radius of a nuclear fireball

Derivation CThe radius of a nuclear fireball

We want $R(E, t, \rho)$ for a strong point explosion, assuming the radius depends only on the energy released, the elapsed time, and the ambient air density. The strategy is equation (1) with ingredients $E$ , $t$ , $\rho$ ; the payoff is inverting the result to extract $E$ from one measured frame of the Trinity film. Watch how the fifth root emerges — and why it makes the estimate robust.

Step 1. Assume $R = k\, E^{a}\, t^{b}\, \rho^{c}$ .

Step 2. Dimensional formulae: $[R] = [\mathrm{L}]$ , $[E] = [\mathrm{M\,L^{2}\,T^{-2}}]$ , $[t] = [\mathrm{T}]$ , $[\rho] = [\mathrm{M\,L^{-3}}]$ .

Step 3. Substitute: $[\mathrm{L}] = [\mathrm{M}^{a+c}\,\mathrm{L}^{2a-3c}\,\mathrm{T}^{-2a+b}]$ .

Step 4. Balance. $[\mathrm{M}]$ : $0 = a + c$ , so $c = -a$ . $[\mathrm{T}]$ : $0 = -2a + b$ , so $b = 2a$ . $[\mathrm{L}]$ : $1 = 2a - 3c = 2a + 3a = 5a$ , so $a = \tfrac15$ , $b = \tfrac25$ , $c = -\tfrac15$ .

Step 5. Assemble: $R = k\left(\dfrac{E t^{2}}{\rho}\right)^{1/5}$ , and inverting for the energy, $E \approx \dfrac{\rho R^{5}}{k^{5} t^{2}}$ with $k$ close to 1 (Taylor's detailed shock calculation gave $k \approx 1.03$ for air). Why the fifth root is forgiving: $E$ scales as $R^{5}$ , so even a generous error in reading $R$ off a grainy photograph changes $E$ by a bounded factor — while conversely $R$ is almost insensitive to $E$ , which is why fireballs of wildly different bombs look deceptively similar at early times.

Step 6. Insert the Trinity frame at $t = 25\ \mathrm{ms}$ , $R \approx 140\ \mathrm{m}$ , $\rho = 1.2\ \mathrm{kg/m^3}$ : $E \approx \dfrac{1.2 \times (140)^{5}}{(0.025)^{2}} = \dfrac{1.2 \times 5.38 \times 10^{10}}{6.25 \times 10^{-4}} \approx 1.0 \times 10^{14}\ \mathrm{J}$ . Verified: $140^5 = 5.38 \times 10^{10}$ ; $1.2 \times 5.38 \times 10^{10} = 6.45 \times 10^{10}$ ; dividing by $6.25 \times 10^{-4}$ gives $1.03 \times 10^{14}\ \mathrm{J}$ . One kiloton of TNT is $4.184 \times 10^{12}\ \mathrm{J}$ , so $E \approx 25\ \mathrm{kt}$ — the declassified yield of Trinity is 21–25 kt.

Where the method fails

Every board that asks you to derive relations also asks where the machine breaks. Each failure traces to one root: equation (1) assumed a single product of powers, and nature is under no obligation to be that simple.

failure mode	example	why it fails
dimensionless constants invisible	the $2\pi$ in $T = 2\pi\sqrt{L/g}$	pure numbers multiply in without disturbing dimensions
sums of terms	$s = ut + \tfrac12 at^{2}$	a product of powers can never produce a sum; check each term instead
more than three unknowns	drag from $\rho$ , $v$ , $A$ , $\eta$	only 3 equations ( $M, L, T$ ); 4 unknowns leave a free parameter
dimensionless variables	amplitude, strain, angle	they enter to any power (or inside any function) undetected
non-power-law forms	$A\sin(\omega t)$ , $N_0 e^{-\lambda t}$	trig/exponential laws are not products of powers
two variables sharing dimensions	two lengths $L_1, L_2$	only their combination is fixed; the ratio $L_1/L_2$ roams free

The deepest failure is the dimensionless one: any function of a dimensionless combination can be smuggled into a "derived" law without disturbing the balance. Dimensional analysis proves necessity, never sufficiency — the same verdict as the checking rule of Part I.

Limiting cases

Case	Result	Meaning
an exponent solves to 0 (mass in the pendulum)	that variable drops out	the method detects irrelevant variables — but only when they carry a unique dimension
unknowns = 3 = equations	unique exponents	the sweet spot: pendulum, string wave, fireball all live here
unknowns > 3	underdetermined system	extra physics (an experiment, a known limit) must fix the surplus
all candidate variables dimensionless	no equations at all	the method is completely silent — pure numbers are its blind spot
both systems share a base unit ( $T_1 = T_2$ )	that ratio is 1	conversion factors simplify — SI↔CGS only touches mass and length

Common trap“Dimensional analysis derives the formula”

It derives the form, never the formula. The pendulum derivation delivers $T = k\sqrt{L/g}$ and stops; the $2\pi$ — a 530% correction over $k = 1$ ! — comes from solving the actual dynamics. Worse, a dimensionally consistent result is not necessarily true: the method happily blesses $T = k\sqrt{L/g}\,f(\theta_0)$ for any function of the dimensionless amplitude — and for large swings the period really does grow; the familiar formula holds only because small-angle physics makes $f \approx 1$ . When an exam asks what dimensional analysis tells us, claim exactly what the bookkeeping proves: the exponents. Claiming the constant, or claiming the law verified, converts a right method into a wrong answer.

Worked examples

Board level — speed of a wave on a stretched string (CBSE/ICSE pattern)

The speed $v$ of a transverse wave on a string is assumed to depend on the tension $F$ in the string and its mass per unit length $\mu$ . Derive the expression for $v$ , and evaluate it for $F = 100\ \mathrm{N}$ and $\mu = 0.01\ \mathrm{kg/m}$ , given that the dimensionless constant is 1.

Step 1. Assume the power law: $v = k\, F^{a} \mu^{b}$ .

Step 2. Dimensions: $[v] = [\mathrm{L\,T^{-1}}]$ , $[F] = [\mathrm{M\,L\,T^{-2}}]$ , $[\mu] = [\mathrm{M\,L^{-1}}]$ .

Step 3. Substitute and collect: $[\mathrm{L\,T^{-1}}] = [\mathrm{M}^{a+b}\,\mathrm{L}^{a-b}\,\mathrm{T}^{-2a}]$ .

Step 4. Balance each dimension. $[\mathrm{M}]$ : $0 = a + b$ . $[\mathrm{T}]$ : $-1 = -2a$ , so $a = \tfrac12$ and hence $b = -\tfrac12$ .

Step 5. Assemble and evaluate: $v = k\sqrt{F/\mu}$ ; with $k = 1$ , $v = \sqrt{100/0.01} = \sqrt{10^{4}} = \boxed{100\ \mathrm{m/s}}$ .

Step 6. The sentence worth a mark: "Dimensional analysis fixes the exponents; $k = 1$ is supplied, since the method cannot determine dimensionless constants." (Full theory says $k$ really is 1 here — but the pendulum's $2\pi$ warns you never to assume that for free.)

Competitive level — a satellite's period from G, M and r (JEE-Main pattern)

Assuming the period $T$ of a satellite in a circular orbit depends on the orbital radius $r$ , the mass $M$ of the planet, and the gravitational constant $G$ ( $[G] = [\mathrm{M^{-1}L^{3}T^{-2}}]$ ), find how $T$ depends on $r$ — and estimate the period of a low-Earth orbit with $r = 6.78 \times 10^{6}\ \mathrm{m}$ , $GM = 3.99 \times 10^{14}\ \mathrm{m^3/s^2}$ , taking the constant as $2\pi$ .

Step 1. Assume $T = k\, G^{a} M^{b} r^{c}$ and substitute dimensions: $[\mathrm{T}] = [\mathrm{M^{-1}L^{3}T^{-2}}]^{a}[\mathrm{M}]^{b}[\mathrm{L}]^{c} = [\mathrm{M}^{-a+b}\,\mathrm{L}^{3a+c}\,\mathrm{T}^{-2a}]$ .

Step 2. Balance. $[\mathrm{T}]$ : $1 = -2a \Rightarrow a = -\tfrac12$ . $[\mathrm{M}]$ : $0 = -a + b \Rightarrow b = -\tfrac12$ . $[\mathrm{L}]$ : $0 = 3a + c \Rightarrow c = +\tfrac32$ .

Step 3. Assemble: $T = k\, \dfrac{r^{3/2}}{\sqrt{GM}}$ — squaring gives $T^{2} \propto r^{3}$ : Kepler's third law falls out of pure dimensions.

Step 4. Evaluate: $r^{3} = 3.12 \times 10^{20}\ \mathrm{m^3}$ ; $r^{3}/GM = 7.81 \times 10^{5}\ \mathrm{s^2}$ ; square root $= 884\ \mathrm{s}$ ; times $2\pi$ : $T = \boxed{5.6 \times 10^{3}\ \mathrm{s} \approx 93\ \mathrm{minutes}}$ — the familiar ~90-minute period of the ISS.

Step 5. MCQ reflex: if the options are exponent patterns ( $T \propto r^{1/2}$ , $r$ , $r^{3/2}$ , $r^{2}$ ), you only need the $[\mathrm{T}]$ and $[\mathrm{L}]$ rows — two lines of algebra, done. Timing note: 90–120 seconds for the full derivation; under 40 seconds if the question only wants the power of $r$ — balance time first (it involves only $a$ ), then length. Solving all three equations for a single-exponent MCQ is donating time to the paper.

Exam traps

Claiming the constant. Writing $T = 2\pi\sqrt{L/g}$ as the output of dimensional analysis loses the reasoning mark. The honest output is $k\sqrt{L/g}$ ; say so, then quote $k = 2\pi$ as supplied by dynamics.

Claiming amplitude independence was proven. The derivation never mentioned amplitude because an angle is dimensionless — the method assumed it away, not disproved it. When boards ask "state one limitation", this is the sharpest answer.

Exponent sign slips. The most common arithmetic death: the $[\mathrm{T}]$ -row gives $1 = -2b$ , and hurried hands write $b = +\tfrac12$ . Anchor: more $g$ must mean faster swings, so $g$ belongs downstairs ( $b < 0$ ). A one-second physical sanity check beats re-solving.

Inverted conversion ratios. In equation (3) the ratios are old-over-new. Test on $1\ \mathrm{m} = 100\ \mathrm{cm}$ (bigger unit → smaller number) before converting anything composite. Half of all dyne/erg errors are a flipped fraction.

Wrong ingredient list. In "assume $Q$ depends on…" questions, use exactly the given list — adding your own variable is not rigour, it is a lost mark, and it usually wrecks the exponent count.

g = 9.8 or 10? CBSE and NEET state which to use per question ( $T$ for a 1 m pendulum: $2.01$ s with $9.8$ , $1.99$ s with $10$ ); JEE-Main options are built from the paper's stated constant. Use exactly what the paper gives.

Origins · 1 minG. I. Taylor and how this was discovered

G. I. Taylor

1886–1975 · Cambridge

Geoffrey Ingram Taylor was the twentieth century's master of small mathematics applied to large physics — turbulence, swimming microbes, crystal dislocations — always with dimensional reasoning first and heavy machinery later. During the war he worked out how a strong point explosion must scale; in 1950, when the Trinity photographs appeared in public, he published the classified yield using nothing but a ruler, the film timestamps, and the fifth-root law you derived above — to the documented irritation of the American security establishment. His papers remain the standard demonstration that knowing what a quantity can depend on is sometimes worth as much as knowing the full theory.

History

1877

— Lord Rayleigh's Theory of Sound systematically uses "the method of dimensions" to guess the forms of physical laws.

1914

— Edgar Buckingham publishes the π-theorem, fixing exactly how many independent dimensionless groups a problem allows.

1941

— Andrei Kolmogorov derives the energy spectrum of turbulence by dimensional reasoning — still the backbone of turbulence theory.

1945

— At the Trinity test, Enrico Fermi estimates the yield within a factor of two by dropping paper scraps into the blast wind.

1950

— G. I. Taylor publishes the blast-wave analysis, extracting the classified Trinity yield from public photographs.

Going further

Texts. NCERT Physics Part 1, Ch. 1 §1.6 — the CBSE/NEET canonical treatment with the pendulum example · H. C. Verma, Concepts of Physics Vol. 1, Ch. 1 · G. I. Taylor, Proc. Roy. Soc. A 201, 175 (1950) — the fireball paper, readable with this article's algebra · Barenblatt, Scaling for where the method leads.

Provenance. All derivations follow the standard power-law method per NCERT Ch. 1. A verifier should recompute: pendulum exponents $c=0$ , $a=\tfrac12$ , $b=-\tfrac12$ and $T = 2\pi\sqrt{1/9.8} = 2.01$ s; $1\ \mathrm{N} = 10^{3+2} = 10^{5}$ dyne and $1\ \mathrm{J} = 10^{3+4} = 10^{7}$ erg; blast exponents $a=\tfrac15, b=\tfrac25, c=-\tfrac15$ and $E = 1.2 \times 140^{5}/0.025^{2} = 1.03 \times 10^{14}$ J $\approx 25$ kt at $4.184 \times 10^{12}$ J/kt; string wave $v = \sqrt{100/0.01} = 100$ m/s; satellite $T = 2\pi\sqrt{3.12\times10^{20}/3.99\times10^{14}} = 5.6 \times 10^{3}$ s.

Prerequisites. Dimensional Analysis I: Checking Equations · Dimensions & Dimensional Formulae. Leads to. Dimensional Shortcuts on MCQs (the technique drill where this method becomes a sub-30-second option-killer) · Propagating Errors: Sums & Differences · Buckingham pi theorem (the UG deep dive).

How does the method know the pendulum period does not depend on mass?

It performs an honest audit of kilograms. The target $T$ carries no mass dimension, and neither $L$ nor $g$ contains any $[\mathrm{M}]$ to cancel against — so the moment you balance the $[\mathrm{M}]$ row, the mass exponent is forced to zero. But notice the fine print: this detection only works because mass carried a unique dimension in the list. A dimensionless quantity like the amplitude slips through the audit entirely — the method is silent about it, which is not the same as proving it irrelevant. Real pendulums do swing slightly slower at large amplitude.

Why exactly can the 2π never be found by dimensions?

Because $2\pi$ is a pure number: multiply any dimensionally balanced equation by it and it stays balanced. The bookkeeping literally cannot tell $T = \sqrt{L/g}$ from $T = 6.283\sqrt{L/g}$ — both are "correct" to the method. The constant must come from somewhere with more information: an experiment (time one swing and divide), or the full equation of motion. This is why mark schemes insist you write $T = k\sqrt{L/g}$ as the derived result and only then quote $k = 2\pi$ as externally supplied.

Why does the method break for s = ut + ½at²?

Two independent reasons, and examiners like both. First, a product of powers $k\,u^{a}t^{b}a^{c}$ can never generate a sum of two terms — the assumed form is structurally wrong. Second, even term by term, the ingredients $u$ , $t$ , $a$ involve only $[\mathrm{L}]$ and $[\mathrm{T}]$ : two equations for three exponents, underdetermined, so $ut$ , $at^{2}$ and stranger products all pass. For multi-term equations, dimensional analysis demotes itself from deriving tool to checking tool — verify each term is a length, as in Part I.

How do I choose which variables Q depends on in the exam?

In board papers you are told — "assuming $T$ depends on $L$ and $g$ " — and you must use exactly that list; adding variables is penalised, not rewarded. In JEE-style questions the list is implied by the setup: ask what the phenomenon can feel. A satellite feels $G$ , the planet's mass and its orbital radius; a fireball feels the energy dumped, the time elapsed, and the density of the air it pushes. If two candidates only ever act as one combination (like $G$ and $M$ in orbits, appearing only as $GM$ ), treat them as one ingredient — a hint from the physics, not a flaw in the method.

Is n₂ = n₁(u₁/u₂) or n₁(u₂/u₁)? I always second-guess the direction.

Anchor it with the one conversion you can never get wrong: $1\ \mathrm{m} = 100\ \mathrm{cm}$ . From metres (big unit) to centimetres (small unit) the number went up, so the value scales as old-unit-over-new-unit: $n_2 = n_1(u_1/u_2)$ — just $nu =$ constant rearranged. For derived units apply this per base dimension with the exponents of the dimensional formula, as in the newton→dyne derivation. Ten seconds re-deriving the direction is cheaper than a wrong power of ten.

If dimensional analysis can produce Kepler's third law, why did we need Newton?

Because the method only converts assumptions into forms. To set up the satellite derivation you already had to know the relevant variables were $G$ , $M$ and $r$ — knowledge that encodes Newton's law of gravitation. And the output, $T^{2} \propto r^{3}$ , arrives with an unknown constant and no explanatory power: it cannot say why orbits are ellipses or connect the Moon to a falling apple. Dimensional analysis is a consistency engine, not a physics generator; it amplifies insight you feed it and produces nothing from an empty tank.

Does dimensional analysis ever use more than three base dimensions?

Yes — the full SI toolkit has seven. Electricity problems regularly need current $[\mathrm{I}]$ as a fourth row (four equations, room for four exponents), and heat problems add temperature $[\Theta]$ . This chapter's mechanics stays inside $[\mathrm{M}]$ , $[\mathrm{L}]$ , $[\mathrm{T}]$ , which is why "three unknowns maximum" is the working rule here. The general accounting — how many independent dimensionless groups survive — is the Buckingham π-theorem, a university topic worth previewing after exam season.