Three arrows make a box
You have climbed the whole rung to reach this corner. Two arrows from a single point, you learned, span a parallelogram, and the cross product u x v hands you both its area (as the length |u x v|) and a perpendicular direction. Now add a third arrow w from the same corner. The three of them, slid tip-to-tail in every order, sweep out a slanted box — a parallelepiped, the three-dimensional cousin of the parallelogram. Six flat faces, all of them parallelograms, opposite ones identical. The question that crowns this rung is the obvious one: *what is the volume of that box?*
The honest way to think about any volume is base times height. Pick two of the arrows, say u and v, to be the floor of the box; that floor is a parallelogram, and you already know its area is |u x v|. The cross product even hands you the floor's normal direction, perpendicular to that floor. The height of the box is then how far the third arrow w reaches straight up out of the floor — not w's own length, but only the part of w that points along the floor's normal. The slant of the box must not count; only the rise off the floor does. That single phrase, 'how much of w points along the floor's normal,' is exactly the job the dot product was built for.
Dot meets cross: the scalar triple product
Put the two pieces together. The floor's area-and-normal live inside one vector, u x v: its length is the floor area, its direction is the floor's normal. The height — w's reach along that normal — is read off by the dot product w . (u x v), since dotting with a vector measures how much of w lies along it. Multiply base by height and you get the volume, and astonishingly the whole computation collapses to a single expression. This number, w . (u x v), is the scalar triple product. One dot, one cross, three vectors, and out drops a number that is the volume of their box.
When you grind out the dot-of-a-cross in components, a clean pattern falls out. With u = (u1, u2, u3), v = (v1, v2, v3), w = (w1, w2, w3), the scalar triple product is the 3-by-3 determinant whose rows are the three vectors. You do not need the theory of determinants to use it — the little expansion below is just bookkeeping — but it is worth seeing, because the determinant view explains every property of the triple product at a glance. Swapping two rows flips its sign; a repeated row makes it zero; scaling a row scales the answer. Each of those algebra facts is about to become a geometry fact.
| u1 u2 u3 |
w . (u x v) = | v1 v2 v3 |
| w1 w2 w3 |
= u1 (v2 w3 - v3 w2) - u2 (v1 w3 - v3 w1) + u3 (v1 w2 - v2 w1)
Volume of the box = | w . (u x v) | (absolute value)
A worked box, and why the sign matters
Take the three unit steps along the axes, u = (1, 0, 0), v = (0, 1, 0), w = (0, 0, 1). Their box is the ordinary unit cube, and the determinant of the identity is 1 — volume 1, exactly as it should be. Now tilt the lid: keep u and v but lean w over to w = (1, 1, 1). The floor area is unchanged at 1, and the triple product is still 1, because volume is base times perpendicular height, and leaning the top arrow sideways does not change how high it rises off the floor. The slanted box has the same volume as the upright cube — this is exactly Cavalieri's principle speaking in vector language: shear a solid sideways and its volume holds.
Now the subtlety the title hinted at. The triple product can come out negative. Volume itself is never negative, which is why the box's volume is the absolute value |w . (u x v)|. So what is the sign telling you? It records orientation — the handedness of the three arrows. If u, v, w form a right-handed set (curl the right hand from u to v and the thumb points the way of w), the product is positive; if they are left-handed, it is negative. The sign is not noise to be discarded; it is genuine information about the twist of the trio, and it is exactly why swapping two of the vectors flips the sign (you have flipped the handedness).
Volume zero means flat: the coplanarity test
Here is the property that earns the triple product its keep. A box has zero volume exactly when it is squashed perfectly flat — when the third arrow w lies in the very plane of the floor spanned by u and v, so the box has no height at all. In symbols: w . (u x v) = 0 if and only if the three vectors are coplanar. The triple product is therefore a one-line coplanarity test: stack the three vectors, take the determinant, and if it is zero they all lie in a single plane through the origin; if it is nonzero they genuinely fill out three dimensions.
Why is this the same as a repeated-row determinant being zero? Because if w can be built out of u and v (which is what 'lying in their plane' means), then the bottom row of the determinant is just a recipe mixed from the rows above it, and any such determinant collapses to zero. The geometry — no height, no volume — and the algebra — dependent rows, zero determinant — are two faces of one fact. This is the everyday tool you reach for to ask 'do these three directions actually escape into space, or are they secretly stuck in a single plane?'
One worked check. Are u = (1, 2, 1), v = (2, 1, 0), and w = (3, 3, 1) coplanar? The third arrow is suspiciously close to u + v = (3, 3, 1) — and indeed it is exactly that. So w is built from u and v, the box is flat, and the triple product must be zero. Compute it and you do get 0, confirming all three lie in one plane through the origin. No drawing, no guessing — just stack, expand, and read the single number.
From the box to every volume
The parallelepiped is the master shape; almost every solid volume you met on the earlier rung is a slice of it. A rectangular box is just the special case where u, v, w are mutually perpendicular, and the triple product reduces to length times width times height — the formula a child knows. Cut the slanted box in half along a diagonal plane and you get a triangular prism, so the prism volume is half the triple product's magnitude. The connection runs both ways: the abstract triple product recovers the homely box formula, and the homely formulas are all shadows of this one expression.
Even the tetrahedron yields. Three arrows from a single vertex of a tetrahedron span a box; the tetrahedron is the corner piece of that box, and it fills exactly one sixth of it. So the volume of the tetrahedron is the triple product's magnitude divided by 6 — the cleanest possible route to the pyramid volume in coordinates. Want the volume of a tetrahedron with given corner points? Form the three edge vectors from one chosen vertex by head minus tail, take their triple product, divide by 6, and you are done. No base area, no height-hunting, no trigonometry.
- Get three edge vectors from one common corner — for points, use head minus tail on the components to build u, v, w.
- Compute the scalar triple product w . (u x v) — most reliably as the 3-by-3 determinant with the three vectors as rows.
- If it is zero, the three are coplanar — flat box, no volume; stop here, you have your answer.
- Otherwise the parallelepiped volume is its absolute value; halve it for a prism, divide by 6 for a tetrahedron, and read the sign as orientation.
Stand back and see what this rung has built. A vector turned an arrow into a list of numbers; the dot product turned angle into a number; the cross product turned area and a perpendicular direction into a vector; and now the triple product folds both together to turn volume, flatness, and handedness into a single number. From a humble arrow you can add, you have climbed to a tool that measures the room. That is the whole promise of vectors made good — and it is the launch pad for the differential-geometry rungs ahead, where these same products, applied to ever-shifting arrows along curves and across surfaces, become curvature, area, and the shape of space itself.