Before you can hunt for those special arrows, you need to be fluent in the alphabet the parent note speaks. Below is every symbol and idea it leans on, built from absolute zero, each one earning its place before the next arrives.
The bold letter v means "the whole arrow", not a single number. The plain numbers 3 and 1 inside are its components — how far right and how far up the arrow reaches.
Why do we need arrows at all? Because everything the parent note does — stretching, rotating, staying-on-a-line — is a statement about directions and lengths, and an arrow is the cleanest picture of a direction with a length.
The special arrow 0=(00) is the zero vector: an arrow of length zero, pointing nowhere. Keep it in mind — it causes the parent note's most subtle rule ("0 is never an eigenvector").
Why introduce this now? Because later ideas — "all combinations of some arrows" (defined properly in Section 12 as span and subspace) — are literally built by stretching and adding arrows. You cannot talk about "all combinations of these arrows" until you can add them, so adding comes first.
Notice: scalar multiplication never rotates the arrow (except the flip at negative λ, which is still on the same line). That is exactly why eigenvectors "stay on their own line."
A matrix is not just storage — it is a function that eats a vector and produces a vector. The rule (matrix–vector multiplication) for a 2×2 is:
Let's watch the machine in action on our example A:
A(31)=(2⋅3+1⋅11⋅3+2⋅1)=(75).
The output arrow (57) points a different way than (13) — the machine rotated it. Most arrows get rotated. The parent note asks: which rare arrows come out pointing the same way, only rescaled? Those are the eigenvectors.
Why do we need I? Here is the short derivation that forces it into existence. We want the non-turning condition Av=λv rewritten as "some matrix times v equals 0", because (as Section 7 shows) that shape is the one we know how to solve.
Hence A−λI makes sense while "A−λ" does not. This is the exact same derivation the parent note performs; we reproduce it here so this page stands alone.
The equation we just derived, (A−λI)v=0, is a homogeneous system: a matrix times an unknown arrow equals the zero vector. ("Homogeneous" just means the right side is all zeros.)
Two arrows always live here for free: 0 itself (every matrix sends 0→0), and possibly a whole line or plane of others.
So the eigenspace Eλ is exactly a null space — which is why, if you know how to find a null space, you know how to find an eigenspace. For the full machinery of solving such systems, see Null space and solving homogeneous systems.
Its meaning in pictures: ∣detM∣ is the area-scaling factor of the machine. If a machine doubles areas, det=±2. If the machine flattens 2D arrows onto a single line (area becomes 0), then det=0 — the machine is called singular.
When you actually compute det(A−λI), the unknown λ appears on the diagonal, so the determinant comes out as a polynomial in λ — powers of λ added together. Let's watch it happen for a general 2×2 so the form is earned, not memorised.
Why is the degree n in general? Because each of the n diagonal entries contributes one factor of (entry−λ), and multiplying n such factors gives a highest power λn. Full details: Characteristic polynomial.
We promised these words back in Section 2; here they are, built from adding (Section 2) and stretching (Section 3).
Why the parent note cares: an eigenspace is not one arrow but a whole subspace of non-turning arrows. Its dimension (the geometric multiplicity, Section 11) is how the parent tells apart Example 2 (a full plane, dim=2) from Example 3 (only a line, dim=1). And because Eλ=Null(A−λI) (Section 7), an eigenspace really is a subspace — a null space always satisfies (a), (b), (c).
R, C ::: the real numbers; the complex numbers (include −1).
λ (lambda) ::: an eigenvalue — a stretch factor.
v (bold) ::: a vector / arrow; bold distinguishes it from a plain number.
v+w ::: vector sum, added component by component (parallelogram tip-to-tail).
0 ::: the zero vector, an arrow of length zero.
A,M ::: matrices (machines that move arrows).
m×n ::: a grid with m rows and n columns; square when m=n.
A−λI ::: matrix subtraction, entry by entry.
I ::: the identity matrix (do-nothing machine).
Eλ ::: the eigenspace for λ — all its non-turning arrows plus 0.
p(λ)=det(A−λI) ::: the characteristic polynomial (degree n).
det ::: determinant, the area-scaling / singularity number.
rank(M) ::: number of independent directions the machine keeps (nonzero rows after row-reduction).
Null(M) ::: null space — arrows crushed to 0.
span{…} ::: all stretch-and-add combinations of the listed arrows.
∈ ::: "belongs to / is a member of" (e.g. v∈Eλ).
Read the map below bottom-up: follow any arrow "X→Y" as "X is needed before Y." Start at the top boxes (raw ideas) and trace down to the single goal box — every path you can walk is a prerequisite you now hold.
The single node at the bottom, Finding eigenspaces, is the parent topic — and it is now fully powered.