Every Isometry Is a Product of Reflections

One mirror, repeated, is enough

In the first two guides of this rung you collected the four rigid motions of the plane — the slide, the turn, the flip, and the glide reflection — and learned to compose them. Now comes the punchline that ties the whole subject together. Of those four, one is secretly the parent of the others. The reflection — the humble flip across a line — is the single building block from which every isometry of the plane can be assembled. That is the claim of the three reflections theorem: every plane isometry is a product of at most three reflections.

Why should you believe this? Start with the simplest case and feel it in your hands. Hold a sheet of paper up to a window and flip it left-to-right across a vertical line: that is one reflection, and the text on the page now reads backwards. Flip it back across the same line and you are home again. But here is the first surprise — flip it across one vertical line, then across a second, parallel vertical line a little to the right. The page is now facing forward again (two flips undid the backwardness), yet it has shifted sideways. Two reflections in a row, across two parallel mirrors, produced a pure slide.

Two parallel mirrors make a translation

Let us make that paper experiment exact, because the numbers are clean and they pin the idea down for good. Put two parallel mirror lines, m and n, a distance d apart, with n to the right of m. Take any point P and reflect it across m to get P_1, then reflect P_1 across n to get P_2. Reflecting across a line keeps a point's height and only changes its left-right position relative to that line, so the whole motion is a horizontal slide. The beautiful fact is how far it slides: P moves exactly 2d to the right — twice the gap between the mirrors — no matter where P started.

mirror m at x = 0,   mirror n at x = d

start:   P  = (x, y)
reflect across m:   P_1 = (-x, y)
reflect across n:   P_2 = (2d - (-x), y) = (x + 2d, y)

net effect:  (x, y) -> (x + 2d, y)   a slide of 2d to the right

Three things in that little computation deserve to be noticed. The starting coordinate x cancels out completely, which is exactly why the result is a genuine translation — every point moves by the same vector. The slide is perpendicular to the mirrors and points from the first mirror toward the second. And the distance is 2d, not d: each reflection contributes its own share. Reverse the order — reflect across n first, then m — and you slide 2d to the left instead. So composing two reflections does not commute; the order of the mirrors decides the direction of travel.

Two crossing mirrors make a rotation

Now tilt the second mirror so the two lines are no longer parallel but cross at a point O, meeting at an angle theta. Reflect across the first line, then across the second. The point O sits on both mirrors, so each reflection leaves it fixed — and a non-identity isometry with a fixed point can only be a rotation about that point (or a reflection, but two reflections together preserve handedness, ruling a flip out). So two crossing mirrors compose to a turn about their intersection O. By exactly the same doubling we saw with parallel lines, the angle of rotation is 2 * theta — twice the angle between the mirrors.

This 'twice the angle' rule is the rotational twin of the 'twice the distance' rule, and together they are the heart of the whole theorem. A quarter-turn (90 degrees) is two reflections in mirrors meeting at 45 degrees; a half-turn (180 degrees) is two reflections in perpendicular mirrors. Both rules also explain a limiting case cleanly: as you slide the crossing point O off to infinity, the two mirrors become parallel and the rotation 'opens up' into a translation. A translation is, in this honest sense, a rotation about a point infinitely far away — slides and turns are not two unrelated motions but two faces of 'two reflections.'

Counting reflections: the parity invariant

Each reflection swaps left and right — it turns a left hand into a right hand, reversing orientation. Compose two reflections and the handedness flips twice, so it is back to normal: translations and rotations preserve orientation, which is why they are called direct isometries. Compose an odd number of reflections and the final result reverses handedness; these are the opposite isometries — a single reflection, or a glide reflection. This gives you a sharp, reliable test you can apply by eye: does the motion turn a figure into its mirror image, or not?

The deep point is that this parity — even or odd number of reflections — is an honest invariant of the isometry itself, not of how you happened to write it. You might decompose one motion into reflections in many different ways, with different mirror lines, but the count is always either always-even or always-odd; you can never write a rotation as three reflections or a glide reflection as two. That is precisely why 'at most three' splits so neatly: even isometries (identity, translation, rotation) take 0 or 2 reflections, and odd isometries (reflection, glide reflection) take 1 or 3. There is no even isometry needing 3 and no odd one needing 2 — parity forbids it.

Three mirrors and the full classification

What about three reflections? An odd count means an opposite isometry, and we have only two of those. If the three mirror lines all pass through a common point or are all parallel, the composition collapses back down to a single reflection. But in the generic case — three lines in general position — the three reflections compose to a glide reflection: a flip across some line followed by a slide along that same line, the very motion you studied in guide 2 as 'footprints in the sand.' So three is genuinely needed; the glide reflection is the one isometry that cannot be built from fewer than three mirrors.

Stand back and the entire landscape snaps into view — this is the classification of plane isometries, and the reflection count organizes it perfectly. Read the table below as a complete census: every rigid motion of the plane, with no exceptions, is exactly one of these four types (counting the identity as a trivial translation), and you can tell which by counting reflections and checking for fixed points. There is nothing else out there. A list that once looked like four loosely-related tricks is revealed as a tidy, finite, fully-understood family.

reflections | orientation | fixed points        | isometry
------------+-------------+---------------------+-------------------
     0      |  preserved  | every point         | identity
     1      |  reversed   | a whole line        | reflection
     2      |  preserved  | one point  (O)      | rotation about O
     2      |  preserved  | none (parallel m,n) | translation
     3      |  reversed   | none                | glide reflection

Why this matters: the algebra under the geometry

Reducing everything to reflections is not just tidy bookkeeping — it changes what you can compute. Composing 'a rotation then a translation then another rotation' directly is a headache, but rewrite each as reflections, then use the freedom to align a shared mirror so that two adjacent reflections cancel, and a long chain melts down to its true short form. The classification table promises that whatever survives must be one of five outcomes, so you always know the shape of the answer before you finish. This is the practical reason mirror-decomposition is a working tool and not just a pretty fact.

Let me be honest about scope, in the spirit of this whole ladder. A fully rigorous proof of the three reflections theorem needs a careful lemma — that an isometry of the plane is completely determined by where it sends any three non-collinear points — and then a short argument lining up those three points with at most three reflections. We have shown you the mechanism faithfully and verified the key cases by hand, but the airtight version belongs to a first proof-based course. The picture you now hold, though, is the true one: reflections generate everything, and the count of them is an honest fingerprint.

One last thread to carry forward. Because reflections generate every isometry, the collection of all isometries forms a single, well-knit algebraic system in which reflections are the generators. That system is a group, and the symmetries of any particular figure — a square, a wallpaper pattern, a snowflake — are exactly the isometries that map it to itself. The very next guide opens that door: it studies the symmetry group of a figure, and you will find that everything there, too, ultimately comes back to mirrors.