Proving Geometry with Vectors
For two thousand years, proving a fact about triangles or parallelograms meant reaching for a
compass, drawing a clever auxiliary line nobody was given, and hoping the right
pair of triangles turned out congruent. Euclid's students spent years learning which line to add
where. Vectors quietly end that hunt: fix an origin
O, and every point A gets a
position vector \vec{a} = \vec{OA}. From there, whole
proofs collapse into algebra — adding, subtracting and scaling vectors — with no
auxiliary construction and no picture required at all.
The single most useful fact follows at once: the vector from
A to B is the difference of their
position vectors,
\vec{AB} = \vec{b} - \vec{a},
because to go from A to B you can travel
back to O along -\vec{a} and then out to
B along \vec{b}. With just this — plus
adding and
scaling
vectors — you can prove facts that are fiddly with pure geometry, and prove them the same
mechanical way every single time.
Three tools
Almost every vector proof is built from three little translations of geometry into vector language:
-
Midpoint. The midpoint M of
AB has position vector
\vec{m} = \tfrac{1}{2}(\vec{a} + \vec{b}) — the average of the ends.
-
Parallel. Two segments are parallel exactly when one vector is a
scalar multiple of the other:
\vec{PQ} = k\,\vec{RS} for some number
k \neq 0.
-
Collinear. Three points A, B, C lie on one straight
line when \vec{AB} = k\,\vec{AC} — parallel vectors that also
share the point A.
"Parallel" and "collinear" look identical in algebra — both are "one is a multiple of the other".
The only difference is that collinear vectors start from a common point, so being
parallel and sharing an endpoint forces the three points onto a single line. Notice, too,
that a vector proof never once measures an angle or a length with a ruler — it reads every fact
straight off an equation.
Worked example 1: the midpoint theorem
Here is a classic result made easy: the segment joining the midpoints of two sides of a
triangle is parallel to the third side and half its length. Take triangle
OAB, and let M, N
be the midpoints of OA and OB. Step through
the argument:
The whole proof is three lines of algebra:
\vec{m} = \tfrac{1}{2}\vec{a}, \qquad \vec{n} = \tfrac{1}{2}\vec{b},
\vec{MN} = \vec{n} - \vec{m} = \tfrac{1}{2}(\vec{b} - \vec{a}) = \tfrac{1}{2}\vec{AB}.
Because \vec{MN} is a scalar multiple
(\tfrac{1}{2}) of \vec{AB}, the two segments
are parallel; and that scalar being \tfrac{1}{2} says
MN is half the length of AB.
Two geometric conclusions, read straight off one vector equation — no auxiliary line, no
protractor, no case-splitting on the shape of the triangle. Whatever shape
OAB has, the algebra doesn't care.
Worked example 2: the diagonals of a parallelogram bisect each other
Give every corner its own position vector from a single fixed origin
O. Let the parallelogram be OACB, with
\vec{OA} = \vec{a} and \vec{OB} = \vec{b} as
its two sides from O. Because opposite sides of a parallelogram are
equal and parallel, the fourth corner is reached by walking along \vec{a}
and then along \vec{b}, so
\vec{OA} = \vec{a}, \quad \vec{OB} = \vec{b}, \quad \vec{OC} = \vec{a} + \vec{b}.
Four corners, four position vectors, all measured from the same O. Now find where the two diagonals cross.
The midpoint of diagonal OC is
\tfrac{1}{2}(\vec{o} + \vec{c}) = \tfrac{1}{2}(\vec{a} + \vec{b}),
and the midpoint of the other diagonal AB is
\tfrac{1}{2}(\vec{a} + \vec{b}) \quad \text{again} — \text{the exact same point.}
Both diagonals pass through \tfrac{1}{2}(\vec{a} + \vec{b}), so they
cross there and nowhere else: the diagonals bisect each other. One shared
expression, computed two different ways, and the theorem falls out — no compass, no congruent
triangles to hunt for.
Worked example 3: the medians of a triangle meet at one point
A median joins a vertex to the midpoint of the opposite side. A triangle has three
of them, one from each vertex — and remarkably, all three always cross at the very same point, the
centroid. Here is the vector argument, outlined rather than run in full: give the
triangle's vertices position vectors \vec{a}, \vec{b}, \vec{c} from an
origin O. The median from A lands on the
midpoint of BC, at \tfrac{1}{2}(\vec{b} + \vec{c}).
The point two-thirds of the way along that median from A is
\vec{a} + \tfrac{2}{3}\!\left(\tfrac{1}{2}(\vec{b}+\vec{c}) - \vec{a}\right) = \tfrac{1}{3}(\vec{a} + \vec{b} + \vec{c}).
Notice this last expression is perfectly symmetric in \vec{a},
\vec{b} and \vec{c} — swap any two labels and
the formula doesn't change. So running the same two-thirds calculation from
B or from C must land on that
identical point. All three medians pass through
G = \tfrac{1}{3}(\vec{a} + \vec{b} + \vec{c}), the centroid — proved
without ever drawing all three medians and squinting to see if they meet.
Proving points collinear
To show three points lie on a line, express two vectors between them and check that one is a
multiple of the other. Say \vec{AB} = 2\vec{p} and, after simplifying,
\vec{AC} = 3\vec{p}. Then
\vec{AC} = \tfrac{3}{2}\vec{AB} — a scalar multiple sharing the point
A — so A, B, C are collinear, and
C is \tfrac{3}{2} of the way along the line
from A through B.
-
Label every point from one fixed origin, and never switch mid-proof. The
moment you write \vec{a} for "position vector of
A from O" partway through, then quietly
reuse \vec{a} to mean something measured from
B instead, the algebra keeps running but the answer is nonsense.
This is the single most common error in vector proofs — pick your origin in the first line and
stick with it to the last.
-
\vec{AB} = \vec{b} - \vec{a}, tip minus tail — not
\vec{a} - \vec{b}. Getting the order backwards flips the vector and
wrecks the proof.
-
"Prove parallel" means showing one vector is a scalar multiple of the other —
not that two lines "look parallel" in a sketch. A picture can mislead; only the
algebra
\vec{PQ} = k\,\vec{RS} counts as a proof. And a scalar multiple only
proves segments parallel — to conclude points are collinear
you also need them to share a point, since parallel alone allows two separate
parallel lines.
For over two thousand years after Euclid, proving a geometry theorem meant inventing a bespoke
picture: draw exactly the right extra line, spot exactly the right pair of congruent triangles,
and hope the trick generalises. Brilliant, but not mechanical — every new theorem needed
a fresh flash of insight. That changed in the 1800s when mathematicians including Hermann
Grassmann and (a little later) Josiah Willard Gibbs built up vector algebra as a tool in its own
right. Suddenly the "flash of insight" step could be replaced by turning a crank: write positions
as vectors, add, subtract, scale, compare. It was a genuine paradigm shift — geometry stopped
being an art of clever pictures and became, for a large class of problems, routine algebra that
even a beginner could carry out correctly.
The same trick is still working for a living today. CAD and robotics software that keeps a bridge
truss rigid, a robot arm's joints consistent, or a 3-D model's parts glued together at the right
points is, underneath, solving exactly these vector equations — midpoints, parallel checks and
all — thousands of times a second, far faster than any human drawing auxiliary lines ever could.