Foundations — Eigenvalues and eigenvectors — characteristic polynomial

Before you can read the parent note, you need to already own a small toolbox of symbols. Below is every symbol and idea the parent uses, built from absolute zero, in an order where each one leans only on the ones before it.

1. The arrow: what a vector $\mathbf{x}$ is

Look at the figure: the red arrow starts at the origin (the point $(0,0)$) and its tip lands at the point $(x_1,x_2)$. The two black dashed legs show the "right" amount $x_1$ and the "up" amount $x_2$.

The special arrow with all zeros, $0=\left(\genfrac{}{}{0ex}{}{0}{0}\right)$, is called the zero vector — it has no length and points nowhere. Remember it: the parent note forbids it as an eigenvector, and we'll see why.

2. The machine: what a matrix $A$ is

Multiplying $A$ by a vector $\mathbf{x}$ produces a new vector $A\mathbf{x}$, computed row-by-row: $A\mathbf{x}=\left(\begin{array}{cc}a& b\\ c& d\end{array}\right)\left(\begin{array}{c}{x}_1\\ x_2\end{array}\right)=\left(\begin{array}{c}a{x}_1+b{x}_2\\ c{x}_1+d{x}_2\end{array}\right).$

In the figure, the black arrow is the input $\mathbf{x}$; the red arrow is the output $A\mathbf{x}$. Notice it points in a different direction — the machine turned it. An eigenvector is the special input where black and red lie on the same line.

3. The stretch number: what a scalar $\lambda$ is

What does each kind of $\lambda$ look like?

• $\lambda =2$: arrow twice as long, same direction.
• $\lambda =\frac{1}{2}$: arrow half as long, same direction.
• $\lambda =-1$: same length, flipped to point the opposite way.
• $\lambda =0$: collapses to the zero vector.

The figure shows one black arrow scaled by $2$, by $-1$, and by $0$ — all three copies stay on the same dashed line through the origin. Staying on that line is the visual meaning of "only stretched, never turned," which is what $A\mathbf{x}=\lambda \mathbf{x}$ demands.

4. The "do-nothing" machine: the identity matrix $I$

Why does the parent note suddenly slip an $I$ into $\lambda I$? Because you cannot subtract a plain number from a matrix: the expression "$A-\lambda$" is meaningless — a grid minus a single number has no defined rule. But scaling the identity, $\lambda I$, is a genuine matrix: $\lambda I=\left(\begin{array}{cc}\lambda & 0\\ 0& \lambda \end{array}\right).$ Now $A-\lambda I$ is a legal grid-minus-grid subtraction. That is the only reason $I$ appears.

5. The volume dial: the determinant $\det$

Why does the topic care about the determinant being zero? A zero area means the parallelogram has been squashed flat onto a line — the machine has crushed 2D down to 1D. When that happens the machine is singular: it destroys information and cannot be undone. See Determinants for the full story.

Left: a normal matrix — its two column arrows (black) span a red parallelogram with real area. Right: a singular matrix — the columns are lined up, the red parallelogram is flattened to a segment, area $=0$.

6. Sending an arrow to nothing: the null space and singular matrices

The parent's Step 3 says: "$M\mathbf{x}=0$ has a nonzero solution iff $M$ is singular." Here is why in plain words:

• If $M$ could be undone (invertible), then applying the undo-machine $M^{-1}$ to both sides of $M\mathbf{x}=0$ gives $\mathbf{x}=0$ — the only solution is the boring zero arrow.
• So a nonzero solution can exist only when $M$ cannot be undone — i.e. $M$ is singular — i.e. $\det M=0$.

That chain, "nonzero solution $\iff$ singular $\iff$ determinant zero," is the engine that turns the eigen-equation into a solvable polynomial. More at Null Space and Rank.

7. The treasure map: what a polynomial and its roots are

For a $2\times 2$ matrix, $\det (A-\lambda I)$ turns out to be a degree-2 polynomial (its highest power is ${\lambda }^2$); for $n\times n$ it has degree $n$. Its roots are the eigenvalues. So the whole hunt reduces to a school-algebra task: find the roots of a polynomial.

8. The diagonal sum: the trace $\mathrm{tr}(A)$

It matters because it equals the sum of the eigenvalues — a one-line check on every answer.

9. When arrows won't cooperate: complex numbers

Why the topic needs them: some real matrices (like a $90^{\circ }$ rotation) turn every arrow, so no real arrow ever stays on its own line. Then the characteristic polynomial, e.g. ${\lambda }^2+1=0$, has no real roots — but it does have the complex roots $\pm i$. Allowing complex $\lambda$ guarantees the treasure map always has treasure.

The prerequisite map

Read it top-down: arrows and machines feed the eigen-equation; the identity lets us form $A-\lambda I$; determinant and null space explain when it's singular; polynomials, trace and complex numbers complete the algebra — and everything pours into the topic.

Equipment checklist

Hide the right side and test yourself. You are ready for the parent note when every line feels obvious.