Vector algebra

The formula

a + b = (a₁+b₁, a₂+b₂, a₃+b₃)

Addition is componentwise, which is just the parallelogram rule written out in coordinates. Two 100-lb forces at right angles sum to 141.4 lb along the diagonal, not 200 lb — the single fact that separates vector arithmetic from the scalar kind.

|a| = √(a₁² + a₂² + a₃²),    â = a / |a|

Magnitude is Pythagoras in three dimensions; dividing a vector by its own magnitude leaves a unit vector — pure direction, length 1 — the standard way to say "along this line" without saying how far.

a · b = a₁b₁ + a₂b₂ + a₃b₃ = |a| |b| cos θ

The dot product: a single number measuring how much of one vector lies along the other. Take a = (3, 4, 0) and b = (4, 3, 0): the left side gives 24, the magnitudes multiply to 25, and acos(24/25) says the vectors sit 16.26° apart. Perpendicular vectors dot to exactly zero — the test engineers use daily without naming it.

a × b = (a₂b₃ − a₃b₂,  a₃b₁ − a₁b₃,  a₁b₂ − a₂b₁),    |a × b| = |a| |b| sin θ

The cross product: a new vector perpendicular to both inputs, right-hand rule picking the side, magnitude equal to the parallelogram area the two vectors span. Same a and b as above cross to (0, 0, −7), and 7 = 25 · sin 16.26° checks.

M = r × F

The moment of a force about a point — the reason mechanical engineers meet the cross product before they meet anything else. A 100-lb force in y applied 2 ft out along x produces (0, 0, 200): 200 ft·lb about z, sign and axis included free of charge.

comp_u(F) = F · û

Projection: the component of F along the unit direction û. Load cells, accelerometer axes, and thrust-along-velocity calculations are all this one line.

Where you meet it

The test stand, resolving load cells. A thrust measurement rig reads six load cells at known orientations; the thrust vector is reconstructed by projecting each reading onto the stand axes and summing componentwise. A 2° gimbal deflection on a 100,000-lbf engine puts 3,490 lbf into the side cells while axial only drops to 99,939 lbf — the sine bleeds fast, the cosine barely moves, and the data-reduction script is nothing but dot products.
The bench, with a tilted accelerometer. The sensitive axis is never quite where the drawing says. The correction is a projection: measured output equals true acceleration dotted with the actual axis direction. Miss that and a 1-g tumble calibration comes back at 0.9994 g and someone chases a phantom scale-factor error for an afternoon.
Statics homework that never ends. Every free-body diagram at every design review is vector algebra: sum of forces zero, sum of moments zero, both componentwise. The bracket margin argument in the review board is settled by whoever resolved the oblique load into axes correctly.
RF and antenna work. Field quantities E and H are vectors; polarization mismatch loss is a dot product between unit polarization vectors, and Poynting flux direction is a cross product. Pointing budgets are unit-vector arithmetic before they are anything else.

How it works

A vector is not "a magnitude and an angle" — that framing works in a plane and collapses in three dimensions. The durable definition is operational: a vector is anything that adds by the parallelogram rule and scales by multiplication. Forces qualify. Velocities qualify. Finite rotations do not — rotate 90° about x then 90° about z, reverse the order, and you end up at a different attitude, so rotation "vectors" don't commute under addition and vector algebra quietly stops applying. That's the boundary of validity most people never get told.

The two products answer different questions and mixing them up is the classic blunder. Dot answers "how aligned?" — scalar out, zero at 90°, maximum when parallel. Cross answers "how much turning, about what axis?" — vector out, zero when parallel, maximum at 90°. Work is F · d; moment is r × F. Write the wrong one in a torque calc and the units still come out N·m, which is exactly why the mistake survives to the review.

Sign conventions do the real damage. The cross product is anticommutative — a × b = −(b × a) — so swapping the order of r and F in a moment calculation flips the direction of every torque in the model and nothing else looks wrong. The right-hand rule is a convention, not a law of nature; it only works if every frame in the analysis is right-handed. One left-handed export from a CAD package (some tools will do this to you on a mirrored part) and the cross products all lie.

Magnitudes don't add. Two 50-lb cables to the same eyebolt can load it at anything from 0 to 100 lb depending on the angle between them; 50 lb and 80 lb at 60° apart resolve to 113.6 lb at 37.6° off the first cable, and the only way to get that number is components. Anyone who adds magnitudes on an oblique load case has built in an error that is sometimes conservative and sometimes fatally not.

Last gotcha: the dot-product angle formula acos(a·b / |a||b|) is numerically ugly for nearly parallel vectors — the cosine flattens near 1 and floating point eats your precision. For small angles between unit vectors, |a × b| (the sine) is the better-conditioned number. Test engineers comparing two nearly aligned sensor axes learn this one empirically.

History

Engineers were adding forces long before anyone called them vectors. Newton put the parallelogram rule at the front of the Principia in 1687 — Corollary I to the laws of motion states that a body acted on by two forces together describes the diagonal of the parallelogram in the same time it would describe the sides separately ^[1]^[2]. For the next century and a half, that geometric rule, plus componentwise bookkeeping, was the whole toolkit.

The algebra arrived by a strange route. On 16 October 1843, walking the Royal Canal in Dublin, William Rowan Hamilton hit on the multiplication rule for quaternions — four-component numbers containing a "vector part" — and carved i² = j² = k² = ijk = −1 into Brougham Bridge on the spot ^[3]^[4]. A year later, working in isolation as a German schoolteacher, Hermann Grassmann published his Ausdehnungslehre (1844), a more general and more powerful system that almost nobody could read; it went largely ignored for decades ^[4]^[5].

The version engineers actually use was distilled from both in the 1880s. Josiah Willard Gibbs at Yale stripped the quaternion down to its useful parts — the dot and cross products as separate operations on plain three-component vectors — and had the notes privately printed for his students in 1881 and 1884 ^[4]^[6]. Oliver Heaviside arrived at essentially the same system independently while rewriting Maxwell's electromagnetism, publishing it in his Electromagnetic Theory from 1893 ^[4]^[7]. The quaternionists did not go quietly: Peter Guthrie Tait, Hamilton's chief advocate, attacked the Gibbs–Heaviside system in print, and the early 1890s saw a running argument in the pages of Nature over which algebra physics should adopt ^[4]^[6]. The fight effectively ended in 1901, when Edwin Bidwell Wilson turned Gibbs's notes into the textbook Vector Analysis; its notation — the dot, the cross, the del — is what's on every whiteboard in the cluster today ^[4]^[6]^[7].

Related tools

/tools/convert-angle — every cos θ and atan2 in a force resolution starts here
/tools/convert-mass-force — get the components into consistent force units before you add them
/tools/friction-force — normal and tangential components are a projection problem first
/tools/projectile-range — initial velocity split into components is the textbook vector decomposition
/tools/bolt-torque — torque is r × F wearing its wrench clothes
/tools/moment-of-inertia-shapes — the next rung up, where moments of vectors become tensors

Sources

Written by HE in our own words from the cited sources — engineering judgment included, your stamp still required. All entries →