Perpendicular Axis Theorem

Some theorems earn their keep by being deep. This one earns it by being free. Take any flat body — a coin, a CD, a sheet-metal bracket, an idealized galaxy disk — and ask for its moment of inertia about the axis poking perpendicularly through it. The perpendicular axis theorem answers: you already know it. It is exactly the sum of the moments about any two perpendicular axes drawn in the plane through the same point, $I_z = I_x + I_y$ . Three numbers that each look like they demand their own integral turn out to carry only two integrals' worth of information; the third comes free of charge. The price of the gift is written in fine print, and the fine print is the real physics of this page. The identity holds because every speck of a lamina — a body of zero thickness — sits at $z=0$ , so Pythagoras splits its distance to the normal axis perfectly between its distances to the two in-plane axes. Give the body any thickness and a spurious $2z^2$ term survives the bookkeeping: for every genuinely three-dimensional object $I_z$ is strictly less than $I_x + I_y$ . A sphere misses the identity by a factor of two. Knowing precisely where flatness enters the proof is what separates using the shortcut from being burned by it. Mastering this article buys you the standard inertia table almost without integration — disk and ring diameter moments in one line, tangent-line moments after a Steiner shift — plus a structural fact about the inertia tensor: a lamina's tensor has a whole row fixed by the other two, which is why every flat body is born a symmetric top.

Key takeaways

For a planar lamina in the $xy$ -plane, the moment of inertia about the normal axis equals the sum of the moments about two perpendicular in-plane axes through the same point: $I_z = I_x + I_y$ .
The proof is one line of Pythagoras: a speck at $(x,y,0)$ sits at distance-squared $x^2+y^2$ from the $z$ -axis, $y^2$ from the $x$ -axis, and $x^2$ from the $y$ -axis — the split is exact speck by speck.
Flatness is load-bearing: for a 3D body the sum $I_x+I_y$ picks up an extra $2\int z^2\,\mathrm{d}m$ , so $I_z < I_x + I_y$ strictly. Never use the theorem on a solid.
Symmetry collapses it further: any lamina with $I_x = I_y$ (disk, ring, square plate) has $I_z = 2I_x$ — diameter moments for free.
Chained with the parallel axis theorem it assembles moments of composite flat shapes with zero new integrals; it also gives the engineering identity $J = I_{xx}+I_{yy}$ for beam cross-sections.
Key idea: for any flat body in the $xy$ -plane, the perpendicular distance-squared to the normal axis is the sum of the perpendicular distance-squareds to the two in-plane axes, so $I_z = I_x + I_y$ .

Where this lives today

The theorem's purest modern habitat is anywhere rotation is engineered from genuinely flat parts. MEMS gyroscopes and resonators — the motion sensors in every phone — are etched silicon laminae a few microns thick, and their designers routinely price in-plane and out-of-plane modes against each other using $I_z = I_x + I_y$ as an exact constraint rather than an approximation, because at a thickness-to-width ratio of $10^{-3}$ the correction term $Mh^2/6$ is parts per million. One step up in scale, reaction wheels for small satellites and the read/write platters in surviving hard drives are spec'd the same way: measure or compute the easy spin-axis moment, and the transverse moments that govern gyroscopic coupling and wobble come from the theorem plus a Steiner shift. The theorem also polices physics engines and robotics simulators from the inside. A rigid body handed to a dynamics engine as an inertia matrix must satisfy the triangle inequalities $I_1 \le I_2 + I_3$ ; the lamina case is the equality boundary, so validators flag any tensor that violates it as unphysical and treat any tensor that saturates it as an infinitely thin sheet — numerically stiff and usually a modeling mistake. And in structural engineering the identity never left: the polar second moment of area of a beam cross-section, $J = I_{xx} + I_{yy}$ , which sets torsional stiffness in every shaft-design handbook, is the perpendicular axis theorem verbatim — exact there for the delightful reason that a cross-section is not approximately flat but flat by definition.

The physical picture

Put a coin flat on the table and look straight down at it. The axis you are looking along — the $z$ -axis, normal to the coin — appears as a single point, the origin. Every speck of the coin is some distance $r$ from that point, and the moment of inertia about the normal axis is the familiar ledger $\int r^2\,\mathrm{d}m$ . Now notice what the top view makes obvious: each speck's position decomposes as $(x, y)$ , and Pythagoras insists $r^2 = x^2 + y^2$ . But $\int y^2\,\mathrm{d}m$ is precisely the moment about the $x$ -axis (for a flat body, a speck's distance to the $x$ -axis is its $y$ -coordinate), and $\int x^2\,\mathrm{d}m$ is the moment about the $y$ -axis. The normal-axis ledger is the other two ledgers added line by line — not approximately, not on average, but speck by speck.

Figure

One speck, three axes - the Pythagorean split that makes the theorem

The cancellation that makes this work hinges entirely on flatness. If the body has thickness, a speck can sit above the plane at height $z$ . Its distance to the $z$ -axis is still $\sqrt{x^2+y^2}$ — climbing parallel to an axis never changes your distance from it — but its distance to the $x$ -axis inflates to $\sqrt{y^2+z^2}$ . The in-plane ledgers now count $z^2$ contributions that the normal-axis ledger never sees, and the tidy split fails in one specific direction: $I_x + I_y$ overshoots $I_z$ . That is why the theorem degrades into the strict inequality $I_z < I_x + I_y$ for every solid, and why the failure is never a surprise sign flip — you always know which way the error points. There is a third way to hold the result, useful for what comes later: as a constraint that reduces three numbers to two. A lamina's rotational identity about a given point in its plane is fully specified by its two in-plane moments; the normal moment is not new information. Symmetry can collapse the count further. A body unchanged by a 90° rotation in its plane — a disk, a ring, a square plate — must have $I_x = I_y$ , and the theorem instantly hands you $I_z = 2I_x$ : one integral now prices all three axes.

The theorem, and where flatness enters

First the cast. Fix an origin $O$ in the body and three perpendicular axes through it. The diagonal moments of inertia are the mass-weighted squared distances to each axis:

I_x=\int (y^2+z^2)\,\mathrm{d}m,\qquad I_y=\int (x^2+z^2)\,\mathrm{d}m,\qquad I_z=\int (x^2+y^2)\,\mathrm{d}m \qquad (1)

where: $\mathrm{d}m$ — the mass element at $(x,y,z)$ relative to $O$ . Each integrand is a squared perpendicular distance to an axis, which is why exactly two coordinates appear in each: the coordinate along the axis is invisible to it. These are the same three diagonal entries that the moment of inertia tensor carries; that article builds the full machine, and this one exploits a special pattern in its diagonal. Now the special assumption. A lamina in the $xy$ -plane is a body whose every mass element satisfies $z=0$ — an idealized flat sheet. Feed that into (1) and the in-plane moments shed their $z^2$ baggage:

z=0\ \text{for all }\mathrm{d}m \;\Longrightarrow\; I_x=\int y^2\,\mathrm{d}m,\qquad I_y=\int x^2\,\mathrm{d}m \qquad (2)

where: the reduction is the whole game — for a flat body, "distance to the $x$ -axis" collapses to the single coordinate $|y|$ . Note what is not assumed: nothing about uniform density, nothing about the shape's outline, nothing about $O$ being the center of mass. Any point of the plane, any in-plane mass distribution. The theorem follows in one line, and the line is worth reading slowly because every symbol is now honest:

I_z = I_x + I_y \qquad \text{(planar lamina; all three axes through one point)} \qquad (3)

where: $I_z$ — the moment about the normal axis; $I_x, I_y$ — the moments about any two perpendicular in-plane axes through the same point. Rotating the in-plane pair about the normal changes $I_x$ and $I_y$ individually, but never their sum — a fact worth flagging now and proving below, because it is the theorem's quiet second gift.

Derivation AGeneral proof and where flatness enters

We are deriving equation (3) from the bare definitions (1), and the point of doing it carefully — rather than waving at Pythagoras — is to isolate the exact step where planarity is consumed, so that the theorem's range of validity is a conclusion rather than a memorized warning label. The strategy is blunt: add the two in-plane moments as defined for a general 3D body, and compare the sum against $I_z$ . Watch the $z^2$ terms: they are the entire story, and the sign of what they leave behind is what turns the failed identity into a strict, one-directional inequality for solids.

Step 1. Write the two in-plane moments for a completely general body and add them:

I_x + I_y = \int (y^2+z^2)\,\mathrm{d}m + \int (x^2+z^2)\,\mathrm{d}m = \int (x^2+y^2+2z^2)\,\mathrm{d}m

Why nothing is assumed yet: this is pure linearity of the integral — the identity so far holds for a potato as much as for a coin. The integrand differs from that of $I_z$ by exactly $2z^2$ .

Step 2. Split off the interloper:

I_x + I_y = \int (x^2+y^2)\,\mathrm{d}m + 2\int z^2\,\mathrm{d}m = I_z + 2\int z^2\,\mathrm{d}m

Why this is already the master statement: the theorem, its failure mode, and the size of the failure are all packed into the one non-negative number $\int z^2\,\mathrm{d}m$ — the mass-weighted spread of the body out of the plane.

Step 3. Impose planarity. For a lamina every element has $z=0$ , so $\int z^2\,\mathrm{d}m = 0$ and

I_z = I_x + I_y

Why this is the only nontrivial step: it is the single point where the hypothesis "flat" is spent. Conversely, for any body with mass genuinely off the plane, $\int z^2\,\mathrm{d}m > 0$ strictly, so $I_z < I_x + I_y$ — the identity never fails upward.

Step 4. Harvest the corollary promised above. Rotate the in-plane axes by any angle about the normal: $I_z$ is manifestly unchanged, and by Step 3 so is the sum $I_x + I_y$ — even though each term shifts individually (that trade is governed by the products of inertia). The sum of in-plane moments is an invariant of the point, not of the axis choice. Verified: reproduces Goldstein §5 and Morin Ch. 8; Step 2's excess term $2\int z^2\,\mathrm{d}m$ evaluates to $Mh^2/6$ for a uniform slab of thickness $h$ (with $\int z^2\,\mathrm{d}m = Mh^2/12$ ), which is the standard thin-plate correction and vanishes as $h\to 0$ , as it must.

Figure

Where flatness enters - and how thickness breaks the identity

Symmetry does the rest

The theorem's daily use is almost always in partnership with a symmetry argument, and the partnership runs one specific way. Suppose the lamina looks the same after some rotation in its own plane that maps the $x$ -axis onto the $y$ -axis — a quarter turn for a square plate, any angle for a disk or ring. Then the two in-plane moments cannot differ, $I_x = I_y \equiv I_\parallel$ , and equation (3) collapses to a two-for-one:

I_z = 2\,I_\parallel \qquad\Longleftrightarrow\qquad I_\parallel = \tfrac12\,I_z \qquad (4)

where: $I_\parallel$ — the moment about any in-plane axis through the symmetry point. The practical direction is the second reading: normal-axis moments are usually the easy integral (all mass at known radius from a point), while diameter moments are the awkward one — and the theorem hands you the awkward one for free. This is the standard derivation of the disk and ring diameter moments, done next, and it is genuinely the fastest route: the direct integral for a disk's diameter moment is a two-variable exercise; the theorem is one division by two.

Figure

Symmetry collapses the theorem - disk and ring diameter moments for free

Go deeper · Derivation B · 1 minDiameter moments of a ring and a disk with no new integrals

Derivation BDiameter moments of a ring and a disk with no new integrals

The goal is the pair of tabulated results $I^{\text{diam}}_{\text{ring}} = \tfrac12 MR^2$ and $I^{\text{diam}}_{\text{disk}} = \tfrac14 MR^2$ , obtained without evaluating a single new integral — the cleanest showcase of the theorem's leverage. The strategy: for each body, write down the normal-axis moment (which is elementary, because every element's distance from the center is known or radially integrable), invoke axisymmetry to equate the two diameter moments, and let equation (4) do the division. The only thing to watch is that "diameter" means an axis through the center, in the plane — the symmetry argument silently uses that both axes pass through the symmetry point.

Step 1. Ring, normal axis. Every element of a thin ring sits at the same distance $R$ from the center, hence from the normal axis:

I_z^{\text{ring}} = \int R^2\,\mathrm{d}m = MR^2

Why this integral is free: the integrand is constant — the ring is the one body whose normal-axis moment needs no calculus at all.

Step 2. Ring, diameters. Axisymmetry makes every diameter equivalent, so $I_x = I_y = I^{\text{diam}}_{\text{ring}}$ , and equation (4) gives

MR^2 = 2\,I^{\text{diam}}_{\text{ring}} \quad\Longrightarrow\quad I^{\text{diam}}_{\text{ring}} = \tfrac12 MR^2

Why the direct route would be uglier: integrating $y^2 = R^2\sin^2\theta$ around the ring works — $\int_0^{2\pi}\sin^2\theta\,\mathrm{d}\theta/2\pi = \tfrac12$ reproduces the same answer — but the theorem replaces the trig identity with a symmetry sentence.

Step 3. Disk, normal axis. Build the disk from rings of radius $r$ and mass $\mathrm{d}m = (2Mr/R^2)\,\mathrm{d}r$ :

I_z^{\text{disk}} = \int_0^R r^2\,\frac{2M r}{R^2}\,\mathrm{d}r = \frac{2M}{R^2}\cdot\frac{R^4}{4} = \tfrac12 MR^2

Step 4. Disk, diameters. Same symmetry, same move:

I^{\text{diam}}_{\text{disk}} = \tfrac12\, I_z^{\text{disk}} = \tfrac14 MR^2

Why this factor of 4 is worth memorizing as a ratio: normal, diameter, and (after Steiner) tangent moments of a disk stand as $\tfrac12 : \tfrac14 : \tfrac54$ — the whole family generated by two theorems from one integral. *Verified: both results match the standard tables (Taylor Ch. 10, Morin Ch. 8); the ring check via $\int\sin^2 = \pi$ agrees; recomputing the disk radial integral gives $2MR^2/4R^2 \cdot R^2$ , i.e. $0.5\,MR^2$ , as stated.*

The lamina's inertia tensor: a row for free

The theorem has a tidy restatement in tensor language, and it explains a fact you will use repeatedly in rigid-body dynamics. For any lamina, both products of inertia involving $z$ die instantly — their integrands carry a factor of $z$ , which vanishes on every element:

I_{xz} = -\int xz\,\mathrm{d}m = 0, \qquad I_{yz} = -\int yz\,\mathrm{d}m = 0 \qquad (5)

where: the minus signs follow the off-diagonal convention of the inertia matrix. The physical content: the normal to a lamina is automatically a principal axis, no diagonalization required — spin a flat body about its normal and $\mathbf{L}$ points along $\boldsymbol{\omega}$ , whatever the shape's outline. If you additionally rotate the in-plane axes to kill $I_{xy}$ (always possible, per the principal-axes eigenproblem), the full tensor of the lamina reads

\mathbf{I} = \begin{pmatrix} I_x & 0 & 0\\ 0 & I_y & 0\\ 0 & 0 & I_x + I_y \end{pmatrix} \qquad (6)

where: the bottom-right entry is not measured but inherited — one full row of the tensor is fixed by the other two. In the language of the tensor article's triangle inequalities ( $I_1 \le I_2 + I_3$ and cyclic), a lamina is the body that saturates one of them: $I_3 = I_1 + I_2$ exactly. Flat bodies live on the boundary of the space of physically realizable inertia tensors — hand a simulation an inertia matrix that violates the inequality and it corresponds to no real object; hand it one that saturates it and you have implicitly declared the body flat. This is also why every axisymmetric lamina is an oblate symmetric top with the ratio locked at $I_z/I_\parallel = 2$ — principal moments $(I_\parallel, I_\parallel, 2I_\parallel)$ , the most oblate a rigid body can be. That locked ratio has a famous dynamical fingerprint, the two-to-one wobble of a tossed plate, unpacked in the FAQ below.

Chaining with Steiner: the composite-shape workflow

The perpendicular axis theorem relates axes through one point; the parallel axis theorem relates parallel axes through different points. Neither alone covers the axis an exam actually asks about — but chained, they span essentially every axis of a flat body, in either order, provided each link's own hypothesis is respected: the perpendicular-axis link needs all three axes through one common point, and the Steiner link needs one of its two axes through the center of mass.

you want	use	notes
$I_z$ from two in-plane moments	eq. (3) at any point of the plane	axes concurrent, lamina only
a diameter moment of a symmetric lamina	eq. (4): $I_\parallel = I_z/2$	the two-for-one; disk, ring, square
an offset in-plane axis	eq. (4), then Steiner $+Md^2$	the chain, done in Derivation C
an offset normal axis	$I_z$ at CM, then Steiner $+Md^2$	no perpendicular-axis step needed
a solid's $I_z$	the tensor, or direct integration	theorem forbidden: $I_z < I_x+I_y$

Go deeper · Derivation C · 1 minChaining the perpendicular and parallel axis theorems

Derivation CChaining the perpendicular and parallel axis theorems

We want the moment of inertia of a uniform disk about a tangent line lying in the disk's own plane — an axis that touches the rim, parallel to a diameter. No single theorem reaches it: it is neither a normal axis nor an axis through the center. The strategy is the standard two-link chain: the perpendicular axis theorem converts the easy normal-axis moment into the central diameter moment, and the Steiner shift then walks that diameter outward by one radius. The order matters and is worth watching — Steiner's hypothesis requires stepping from the center-of-mass axis, so the perpendicular-axis link must run first.

Step 1. Central diameter moment, from the previous derivation's two-for-one:

I_{\text{diam}} = \tfrac12\,I_z = \tfrac12\cdot\tfrac12 MR^2 = \tfrac14 MR^2

Step 2. Steiner shift to the tangent. The tangent line is parallel to a diameter and displaced by the perpendicular distance $d = R$ :

I_{\text{tangent}} = I_{\text{diam}} + MR^2 = \tfrac14 MR^2 + MR^2 = \tfrac54 MR^2

Why the shift term dominates: the displaced-axis penalty $Md^2$ carries the whole mass at the full displacement, while the shape's own spread about its center contributes only the smaller $\tfrac14 MR^2$ — a pattern typical of tangent axes.

Step 3. Cross-check with the theorem applied at the rim point. Equation (3) holds at any point of the plane, so at the tangency point $P$ the normal-axis moment must equal the sum of the two in-plane ones there. The in-plane axes through $P$ : the tangent line itself ( $\tfrac54 MR^2$ , Step 2) and the radial line through $P$ — which is the same line as an extended diameter, so its moment is $\tfrac14 MR^2$ . Sum: $\tfrac54 + \tfrac14 = \tfrac32\,MR^2$ . Independently, Steiner on the normal axis gives $I_z(P) = \tfrac12 MR^2 + MR^2 = \tfrac32 MR^2$ . Why this closure is a real test and not circular: the two routes to $\tfrac32 MR^2$ use the chain in different orders and different links — agreement confirms both theorems and their hypotheses simultaneously. This is exactly the consistency audit worth running on any composite-shape calculation. *Verified: $\tfrac54 MR^2$ matches the standard tables for a disk about an in-plane tangent; the rim-point closure $\tfrac54 + \tfrac14 = \tfrac32$ agrees with the Steiner value of the normal axis at the rim.*

Limiting cases

Case	Result	Meaning
in-plane symmetric lamina	$I_x = I_y \Rightarrow I_z = 2I_x$	disks, rings, square plates, regular polygons: one integral prices all three axes
thin rod along $x$	$I_x \to 0,\; I_z = I_y$	a 1D rod is a degenerate lamina; normal and perpendicular in-plane moments coincide
slab of thickness $h$	$I_x + I_y = I_z + \tfrac16 Mh^2$	the excess $2\int z^2\mathrm{d}m$ quantified; relative error in the theorem $\approx h^2/3R^2$ for a disk
general 3D solid	$I_z < I_x + I_y$ strictly	equality iff planar; a solid sphere misses by a factor of 2 ( $\tfrac25$ vs $\tfrac45\,MR^2$ )
any 3D body, trace form	$I_x+I_y+I_z = 2\int r^2\mathrm{d}m$	the only relation that survives in 3D — a trace identity, not a shortcut

Common trapIt has perpendicular axes, so the theorem applies

The theorem's name advertises the geometry of the axes, but its hypothesis is about the body: it must be a plane lamina, with the two summed axes in that plane and all three axes through one common point. Applying $I_z = I_x + I_y$ to a solid — a sphere, a cube, a cylinder of any finite thickness — is the classic error, and it always errs in the same direction: the true $I_z$ is smaller, by exactly $2\int z^2\,\mathrm{d}m$ . For a solid sphere the "theorem" predicts $I_z = \tfrac45 MR^2$ against the true $\tfrac25 MR^2$ — off by a factor of two. The second, subtler trap is the common-point requirement: summing the moment about one axis through the center with the moment about a perpendicular axis through the rim relates nothing to nothing. Check flatness first, then concurrency — then compute.

Worked examples

1 · A CD, three axes priced from one formula (confidence builder)

A compact disc has mass $M = 16\ \mathrm{g}$ and radius $R = 6.0\ \mathrm{cm}$ . Treating it as a uniform lamina, find its moment of inertia about the spin axis and about a diameter, and the kinetic energy stored at $8{,}000\ \mathrm{rpm}$ (a fast optical drive).

Step 1. Spin (normal) axis, from the disk formula: $I_z = \tfrac12 MR^2 = \tfrac12 (0.016)(0.060)^2 = \mathbf{2.88\times10^{-5}\ kg\,m^2}$ .

Step 2. Diameter, from the theorem's two-for-one (4): $I_{\text{diam}} = \tfrac12 I_z = \mathbf{1.44\times10^{-5}\ kg\,m^2}$ — no integral, just the symmetry sentence " $I_x = I_y$ " and a division.

Step 3. Energy at $8{,}000\ \mathrm{rpm}$ : $\omega = 8000 \times 2\pi/60 = 838\ \mathrm{rad/s}$ , so $T = \tfrac12 I_z\omega^2 = \tfrac12(2.88\times10^{-5})(838)^2 = \mathbf{10.1\ J}$ — the kinetic energy of a cricket ball lobbed at $11\ \mathrm{m/s}$ , stored in 16 grams of spinning polycarbonate.

Step 4. Audit the lamina idealization. The disc is $1.2\ \mathrm{mm}$ thick, so the theorem's error term is $\tfrac16 Mh^2$ relative to $I_z = \tfrac12 MR^2$ : fractional error $h^2/3R^2 = (0.0012)^2/(3\times0.060^2) = 1.3\times10^{-4}$ — about $0.013\%$ . The lamina is an excellent model. (The 15-mm center hole is a bigger effect: the annulus formula $\tfrac12 M(r^2+R^2)$ with $r=7.5\ \mathrm{mm}$ raises $I_z$ by $1.6\%$ for the same mass.)

2 · A washer about a tangent line, with a rim-point closure test (exam level)

A flat steel washer has mass $M = 60\ \mathrm{g}$ , inner radius $R_1 = 2.0\ \mathrm{cm}$ , outer radius $R_2 = 4.0\ \mathrm{cm}$ . Find its moment of inertia about a line in its plane, tangent to the outer rim — then verify the answer by applying the perpendicular axis theorem at the tangency point.

Step 1. Normal axis through the center. For an annulus, $I_z = \tfrac12 M(R_1^2 + R_2^2) = \tfrac12(0.060)(0.020^2 + 0.040^2) = \tfrac12(0.060)(0.0020) = \mathbf{6.0\times10^{-5}\ kg\,m^2}$ .

Step 2. Central diameter, by the theorem with axisymmetry: $I_{\text{diam}} = \tfrac12 I_z = \mathbf{3.0\times10^{-5}\ kg\,m^2}$ .

Step 3. Steiner shift to the tangent line (parallel to a diameter, displaced $d = R_2 = 0.040\ \mathrm{m}$ ): $I_{\text{tan}} = I_{\text{diam}} + MR_2^2 = 3.0\times10^{-5} + (0.060)(0.0016) = 3.0\times10^{-5} + 9.6\times10^{-5} = \mathbf{1.26\times10^{-4}\ kg\,m^2}$ .

Step 4. Closure test at the rim point $P$ . The theorem holds at any point of the plane, so at $P$ : $I_z(P) = I_{\text{tan}} + I_{\text{radial}}$ , where the radial line through $P$ is the same line as an extended diameter, $I_{\text{radial}} = 3.0\times10^{-5}$ . Sum: $1.26\times10^{-4} + 3.0\times10^{-5} = 1.56\times10^{-4}\ \mathrm{kg\,m^2}$ . Independent route — Steiner on the normal axis: $I_z(P) = I_z + MR_2^2 = 6.0\times10^{-5} + 9.6\times10^{-5} = 1.56\times10^{-4}\ \mathrm{kg\,m^2}$ . The closure holds exactly — both theorems, both hypotheses, confirmed in one line.

Step 5. Reading the numbers: the Steiner penalty $MR_2^2 = 9.6\times10^{-5}$ dwarfs the washer's own central diameter moment — for tangent axes, where the axis sits matters roughly three times more than what the shape is.

Origins · 1 minChristiaan Huygens and how this was discovered

Christiaan Huygens

1629–1695 · The Hague and Paris

The mathematics of moments of inertia was born a century before the term existed, in Huygens's struggle with a brutally practical question: why does a real pendulum — a swinging church lamp, a clock's compound bob — not beat time like the idealized point mass? His Horologium Oscillatorium (1673), written to perfect the pendulum clock he had invented in 1656, solved the compound pendulum by introducing exactly the quantity $\sum m_i r_i^2$ and the notion of a center of oscillation, effectively computing moments of inertia decades before Euler formalized them and long before the lamina identities were codified. Huygens was the complete natural philosopher of his generation: he resolved Saturn's rings and discovered Titan with lenses he ground himself, built the first working pendulum clock, derived the formulas of centripetal force independently of Newton, and championed the wave theory of light whose principle still bears his name. The perpendicular axis theorem itself is a later, anonymous distillation — a one-line corollary of the integral definitions — but the intellectual machinery it lives inside, mass-weighted squared distances as the currency of rotation, is Huygens's coinage. He spent his last years back in The Hague, publishing the Traité de la Lumière and speculating, in Cosmotheoros, about life on other worlds.

History

1673

— Huygens's Horologium Oscillatorium computes compound-pendulum periods via mass-weighted squared distances, the first working use of moments of inertia.

1749

— d'Alembert and Euler's correspondence on rigid bodies sharpens the distinction between axes fixed in space and axes fixed in the body.

1765

— Euler's Theoria motus corporum solidorum builds the full apparatus of moments about arbitrary axes, principal axes included; the lamina identities become one-line corollaries.

1834

— Poinsot's geometric mechanics recasts inertia as an ellipsoid, where a lamina's saturated triangle inequality marks the boundary of realizable bodies.

1855

— Beam theory in the hands of Saint-Venant institutionalizes the cross-section version, the polar second moment J equals Ixx plus Iyy, which engineers still use daily.

Going further

References. Morin, Introduction to Classical Mechanics, Ch. 8 (Angular Momentum, Part II) · Goldstein, Poole & Safko, Classical Mechanics, 3rd ed., Ch. 5 · Landau & Lifshitz, Mechanics, 3rd ed., §32 · Taylor, Classical Mechanics, Ch. 10

Provenance. Equations derived from the standard integral definitions of the diagonal moments of inertia; the planar restriction follows from $\int z^2\,\mathrm{d}m=0$ . Standard numbers (disk $\tfrac12MR^2$ , $\tfrac14MR^2$ ; rectangular plate $\tfrac1{12}M(a^2+b^2)$ ; ring $MR^2$ ; tangent-line $\tfrac54MR^2$ ) are textbook values. A verifier should recompute the one-line proof and the disk/plate/ring moments against Morin Ch. 8, Goldstein §5, Landau-Lifshitz §32, and Taylor Ch. 10. Worked-example numbers on this page (CD: $I_z = 2.88\times10^{-5}$ , $I_{\text{diam}} = 1.44\times10^{-5}\ \mathrm{kg\,m^2}$ , $T = 10.1\ \mathrm{J}$ at 8,000 rpm; washer: $6.0\times10^{-5}$ , $3.0\times10^{-5}$ , $1.26\times10^{-4}$ , closure $1.56\times10^{-4}\ \mathrm{kg\,m^2}$ ; slab correction $Mh^2/6$ ) are illustrative arithmetic recomputed from the boxed formulas.

Builds on: Moment of Inertia Tensor · Products of Inertia and the Inertia Matrix Leads to: Parallel Axis (Steiner) Theorem · Principal Axes and Principal Moments of Inertia · Classification of Rigid Bodies: Spherical, Symmetric, and Asymmetric Tops · Inertia Ellipsoid

Does the common point have to be the center of mass?

No — and this is worth internalizing, because it unlocks the closure tests used in this article. The proof of $I_z = I_x + I_y$ never mentions the center of mass: it needs only that all three axes pass through one point and that every element has $z=0$ relative to the lamina's plane. The theorem therefore holds at the rim of a disk, at the corner of a plate, anywhere. What does care about the center of mass is the parallel axis theorem — its shift $Md^2$ is only valid stepping from a CM axis. In chained calculations the division of labor is exactly that: perpendicular-axis works at whatever point is convenient; Steiner is the link that must route through the CM.

Why is it always I_z less than I_x plus I_y for solids — could a weird shape overshoot the other way?

No shape can. The exact identity for any body is $I_x + I_y = I_z + 2\int z^2\,\mathrm{d}m$ , and the correction term is an integral of a square times a positive mass density — it is non-negative for every conceivable object and zero only when all mass sits at $z=0$ . So the failure of the theorem in 3D is one-directional by construction: the in-plane sum always overshoots. This is also the fastest sanity check on a computed or simulated inertia tensor: if someone hands you diagonal moments with $I_z > I_x + I_y$ about any axis ordering, the tensor corresponds to no physical body — the triangle inequality it violates is this same statement dressed in principal-frame clothing.

Is there a three-dimensional version of the theorem?

Not a two-term one. The only exact relation that survives for a general solid is the trace identity: adding all three diagonal moments gives $I_x + I_y + I_z = 2\int r^2\,\mathrm{d}m$ , where $r$ is the distance from the origin (not from an axis). It is a genuine constraint — it says the trace of the inertia tensor is twice the polar moment about the point — but it relates three numbers to one integral rather than fixing one moment from the other two, so it has far less computational leverage. The perpendicular axis theorem is the special magic of $z=0$ : planarity converts the trace identity's three-way share-out into a clean two-plus-one split.

My plate is 2 mm thick, not zero. How wrong is the theorem, actually?

Quantifiably, and usually negligibly. For a uniform slab of thickness $h$ , the out-of-plane spread is $\int z^2\,\mathrm{d}m = Mh^2/12$ , so the exact statement becomes $I_x + I_y = I_z + Mh^2/6$ . Relative to a disk's $I_z = \tfrac12 MR^2$ , the fractional error in using the lamina theorem is $h^2/(3R^2)$ . For the worked example's CD ( $h = 1.2\ \mathrm{mm}$ , $R = 60\ \mathrm{mm}$ ) that is $0.013\%$ — far below manufacturing tolerances. The rule of thumb: the lamina idealization is safe whenever thickness is under a tenth of the lateral size (error under $0.4\%$ ), and the error always makes the true $I_z$ slightly smaller than $I_x + I_y$ predicts.

Do the two in-plane axes have to be symmetry axes or principal axes?

No — any two perpendicular in-plane lines through the common point work, symmetric or not, principal or not. The proof only ever uses $x^2 + y^2 = r^2$ , which holds for every orthogonal pair. There is a bonus hiding here: since $I_z$ obviously does not care how you orient the in-plane pair, the sum $I_x + I_y$ is invariant under rotations of the axes about the normal, even though each term changes individually (their trade-off is tracked by the product of inertia $I_{xy}$ ). That invariance is the 2D trace theorem, and it is occasionally the fastest route to an awkward moment: compute the sum in whatever orientation is easy, subtract the one in-plane moment you know.

Engineers write J equals Ixx plus Iyy for beam cross-sections. Is that the same theorem?

The same identity, one abstraction over. Beam torsion theory uses the second moment of area — the same integrals with area elements $\mathrm{d}A$ in place of mass elements $\mathrm{d}m$ — and the polar second moment $J = \int r^2\,\mathrm{d}A$ that sets torsional stiffness decomposes as $J = I_{xx} + I_{yy}$ by exactly the Pythagorean split of this article. What is pleasing is that in this application the theorem is not an idealization at all: a cross-section is a genuine plane region, flat by definition rather than by approximation, so the identity is exact. One caution carries over: $J = I_{xx}+I_{yy}$ gives the geometric polar moment; equating it to the torsional stiffness constant is itself only exact for circular sections — for other shapes warping reduces the effective constant below $J$ .

Why does a tossed plate wobble at twice its spin rate, and what does that have to do with this theorem?

The factor of two is the theorem wearing dynamics clothing. A flat axisymmetric body has principal moments $(I_\parallel, I_\parallel, 2I_\parallel)$ — the ratio $I_z/I_\parallel = 2$ is fixed by $I_z = I_x + I_y$ with $I_x = I_y$ , before any motion is considered. Feed that ratio into torque-free rigid-body dynamics and the wobble (the precession of the symmetry axis seen from outside) comes out at very nearly twice the spin rate for small tilt: the rate involves $I_z/I_\parallel$ , and $2$ is what the lamina geometry serves. This is the wobbling cafeteria plate whose "two-to-one" puzzled Feynman into re-deriving rigid-body precession. The full story — Euler's equations, body versus space frames — belongs to the angular momentum and inertia ellipsoid articles downstream; the ratio they start from is settled here.

Perpendicular Axis Theorem

Where this lives today

The physical picture

The theorem, and where flatness enters

Symmetry does the rest

The lamina's inertia tensor: a row for free

Chaining with Steiner: the composite-shape workflow

Limiting cases

Worked examples

1 · A CD, three axes priced from one formula (confidence builder)

2 · A washer about a tangent line, with a rim-point closure test (exam level)

Going further

Before this

Where this leads