Scalars, Points, Vectors, and Blades Across Dimensions

For new readers

The rest of this series assumes you already know what a vector is, what a basis looks like, and what it means for two vectors to span something. This page is the page before that one. It doesn't assume any of it — only that you can picture a number line. Read this first if a sentence like "a vector is a directed quantity, not a point" has ever left you unsure which of those two things you were actually looking at; everything else in the series builds on the distinction this page exists to make clearly, once, from scratch.

Here is a question that sounds too simple to be worth asking: what does the number 3 mean? In one sense nothing could be more obvious — it's three. But put it on a line, and "3" starts doing at least three different jobs, and geometric algebra cares, a great deal, about which job is in play at any given moment. This page builds that distinction up one dimension at a time, then shows the pattern that keeps it organised no matter how many dimensions you add. the blade is the geometric object

One Number, Three Meanings

Three is a scalar: a plain number, with no geometry attached to it at all. It doesn't live anywhere, and it doesn't point anywhere. Multiply two scalars and you get a scalar back — nothing about that calculation knows whether you're counting apples or measuring time.

As a point, though, three is a specific place on a line: three units from wherever you've agreed to call zero, in whichever direction you've agreed to call positive. A point needs that agreement to mean anything at all — "the point at 3" is meaningless until an origin and a direction have been fixed first. Move the origin, and the same physical spot on the line is now called something else; the point didn't move, but its number did.

And as a vector, three is a displacement of length 3, in a chosen direction, with no origin involved at all. A vector doesn't live at a place — it describes how far and which way to go from any starting point. The same vector drawn starting at one spot and drawn starting at a completely different spot is the same vector both times, which is exactly the property a point does not have.

Fig. 0.1. The same numeral, three different objects. Only the point needs an origin; only the vector is indifferent to where it starts.

The test that tells them apart. Ask: does this number change if I move the origin? A scalar's doesn't, because it never depended on one. A point's does — that's the whole reason it needed an origin in the first place. A vector's doesn't either, for a different reason: it was never measured from the origin to begin with, only from wherever it happens to start.

Two Dimensions: Vectors Need Two Numbers, and a New Object Appears

Move up to the plane, and scalars don't change at all — a scalar is a scalar whether you're working in one dimension or three hundred, because it never had any dimension attached to it. Points and vectors do change: a point now needs two coordinates, and so does a vector. The playground below shows a single vector a both ways at once — as two components, and as a length and an angle — so you can see they're two descriptions of the same object, not two different objects.

Try it yourself (interactive version loads if your browser runs JavaScript): drag the vector's tip and watch its two component readouts — "ax e1" along the horizontal axis and "ay e2" along the vertical — change together with its length and angle. All four numbers describe the same arrow.

What You Can Do With a Vector

Two operations come for free once a vector has components: you can add two vectors together, and you can scale one by a plain number. Adding a and b means placing b's tail at a's tip and drawing from the origin to where b now ends up — tip-to-tail placement is not a convention chosen for convenience, it's what addition is, geometrically. Scaling a vector by a number λ stretches or shrinks it, reverses it when λ is negative, and collapses it to nothing at all when λ is exactly zero.

Try it yourself (interactive version loads if your browser runs JavaScript): the dropdown switches between adding two vectors (drag either one, and toggle between the tip-to-tail and parallelogram constructions) and scaling one vector by a slider λ, including dragging λ negative or down to zero.

Splitting a Vector in Two

Given a reference direction, any vector a splits cleanly into two parts: a projection, the part of a that lies along the reference direction, and a rejection, whatever is left over once that part is removed. The rejection is always perpendicular to the reference direction — not approximately, exactly, for every vector and every reference direction, which the playground computes and checks live rather than asking you to take on faith. this is the dot product and the perp complement

Try it yourself (interactive version loads if your browser runs JavaScript): drag vector a and watch it split into a projection (along the dashed reference line) and a rejection (the right-angle piece left over), which always add back up to a exactly.

A New Kind of Object: The Bivector

Here is where the plane offers something one dimension never could. Two vectors a and b, drawn from the same origin, span a parallelogram between them. That parallelogram has an area, and it has an orientation — sweeping from a round to b goes either counter-clockwise or clockwise, and swapping which vector is which reverses it. The object that carries both the area and the orientation together is called a bivector, and the operation that produces it from two vectors is the wedge product, written a ∧ b.

A bivector has no one-dimensional analogue at all — there was nothing in the previous section waiting to be generalised into it. That's the pattern to notice: each new dimension doesn't just give existing objects more room, it introduces genuinely new kinds of object that didn't exist before.

Try it yourself (interactive version loads if your browser runs JavaScript): drag either vector and watch the shaded parallelogram, its signed area, and the small curved orientation arrow all update together. Bring the two vectors into line with each other and the parallelogram — and the wedge product — collapses to nothing.

Not a rotation, and not just a number. A bivector is sometimes drawn as a curved arrow, which can make it look like a rotation. It isn't one — it's a flat patch of oriented area, a static geometric object in its own right, grade 2 in exactly the sense that a vector is grade 1. A later page in this series, Vectors, Quaternions, and Blades, is where bivectors start generating rotations; that's a genuinely separate idea built on top of this one, not a restatement of it.see how bivectors become rotations there

Three Dimensions: New Objects Again

Add a third dimension and the same two things happen again, at the next grade up. Points and vectors now need three coordinates instead of two. And a new object appears that had no analogue in the plane: the trivector, built from three vectors, representing an oriented volume rather than an area. Three vectors that happen to lie in a single plane span no volume at all — the 3D analogue of two parallel vectors wedging to zero in 2D. volume is a signed box

Bivectors don't disappear in 3D; there are simply more of them available. In the plane there was exactly one independent oriented-area direction to have. In 3D there are three: the e₁₂ plane, the e₁₃ plane, and the e₂₃ plane — any bivector in 3D is some combination of those three. The second example below was built for a different page's discussion of rotors, but its "Plane" control is, for this page's purposes, a direct way to look at each of those three bivector directions in turn: set the angle to whatever you like and just step the plane through xy, yz, and zx.

Try it yourself (interactive version loads if your browser runs JavaScript): the first example sweeps a third vector c out of the plane of a and b and prints the resulting trivector's volume, which goes to zero exactly when all three vectors are coplanar. Switch to the second example and step through the Plane control (xy, yz, zx) to see each of 3D's three independent bivector directions in turn.

The Pattern: Counting the Grades

Collect what's happened so far into one table, and a clean rule falls out of it.

Grade	Object	Count in 1D	Count in 2D	Count in 3D
0	Scalar	1	1	1
1	Vector	1	2	3
2	Bivector	—	1	3
3	Trivector	—	—	1

The number of independent grade-k directions in n dimensions is the binomial coefficient C(n,k) — the same count that answers "how many ways can I choose k things out of n?", because a grade-k blade's basis direction is a choice of k basis vectors out of the n available, wedged together. In 3D that gives the row 1, 3, 3, 1 above, which sums to 8 — the 2³ total basis elements of a multivector in three dimensions. In 4D the same rule gives 1, 4, 6, 4, 1; the pattern keeps going exactly the same way for as many dimensions as a problem actually needs, with no new idea required to get there, only more terms in the same count.

The Geometry of Doubt: Blades as Spatial Boxes — Chapter 1 of the series this page is the on-ramp to, starting from exactly the grade and blade vocabulary built here.
Vectors, Quaternions, and Blades: Why Geometric Algebra Subsumes Both — where the bivector introduced here starts generating rotations, and where this page's grade-counting table is put to work proving a structural claim.
Building Shapes from Blades — the trivector introduced here, built into points, lines, planes and volumes step by step.
Quadrics and the Limits of the Blade — uses this page's grade-counting pattern to explain why a higher-grade blade doesn't gain the extra coefficients a general quadric needs.