Eigenvalue & Eigenvector Calculator

Definition of eigenvalues and eigenvectors.

Let $A$ be a square matrix. An eigenvector is a nonzero vector $v$ for which $Av=\lambda v$ for some scalar $\lambda$. The number $\lambda$ is the corresponding eigenvalue. The condition is simple but powerful: the matrix transforms the vector without changing its line through the origin. It may stretch the vector, shrink it, or reverse its direction, but the vector stays on the same line.

That is why eigenvectors are often described as invariant directions. For a 2×2 matrix, most vectors are pushed into new directions by shear, stretch, or rotation. By contrast, an eigenvector line survives the transformation. If $\lambda >1$, the direction expands. If , the direction contracts. If , the vector is scaled and flipped. When the eigenvalues are complex, the plane still has meaningful behavior, but not along real invariant lines.

Characteristic polynomial for a 2×2 matrix.

Start with the general matrix $A=\left(\begin{array}{cc}a& b\\ c& d\end{array}\right)$. To find eigenvalues, solve the characteristic equation $\det (A-\lambda I)=0$, where $I$ is the identity matrix. For this 2×2 case, $A-\lambda I=\left(\begin{array}{cc}a-\lambda & b\\ c& d-\lambda \end{array}\right)$.

Taking the determinant gives $\det (A-\lambda I)=(a-\lambda )(d-\lambda )-bc$. Expanding the expression produces the quadratic equation ${\lambda }^2-(a+d)\lambda +(ad-bc)=0$. That quadratic is the characteristic polynomial, and its roots are precisely the eigenvalues.

It is often cleaner to rewrite the polynomial using the trace and determinant. Since $\mathrm{tr}\left(A\right)=a+d$ and $\det \left(A\right)=ad-bc$, the same equation becomes ${\lambda }^2-\mathrm{tr}\left(A\right)\lambda +\det \left(A\right)=0$. The discriminant $\Delta ={\mathrm{tr}\left(A\right)}^2-4\det \left(A\right)$ tells you what kind of roots to expect, so it also tells you what kind of geometry the matrix can have.

What trace and determinant tell you before you solve.

Even before computing exact eigenvectors, the identities $\mathrm{tr}\left(A\right)={\lambda }_1+{\lambda }_2$ and $\det \left(A\right)={\lambda }_1{\lambda }_2$ give useful intuition. The trace is the sum of the eigenvalues, so it measures the overall tendency of the system toward growth or decay. The determinant is the product of the eigenvalues, so it captures area scaling and orientation. A negative determinant means the transformation reverses orientation. A determinant of zero means one eigenvalue is zero, which forces all vectors in some direction to collapse.

These quantities also give quick stability clues. In a discrete system such as $x_{n+1}=A{x}_n$, repeated multiplication behaves very differently when than when $\left|\lambda \right|>1$. Values inside the unit circle tend to damp out. Values outside it tend to grow. The calculator does not replace a full systems analysis, but it gives the first facts you need in order to reason about long term behavior.

How eigenvectors are computed.

Once an eigenvalue $\lambda$ is known, the next step is to solve $(A-\lambda I)v=0$. For the 2×2 matrix above, this means solving $\left(\begin{array}{cc}a-\lambda & b\\ c& d-\lambda \end{array}\right)\left(\begin{array}{c}x\\ y\end{array}\right)=\left(\begin{array}{c}0\\ 0\end{array}\right)$. Because the determinant is zero at an eigenvalue, the two equations are dependent rather than fully independent. That leaves a line of nonzero solutions.

In practice, you choose one convenient component and solve for the other. For example, you might set $x=1$ and then solve for $y$. Any nonzero scalar multiple describes the same eigendirection, which is why eigenvectors are not unique. A calculator often reports a normalized vector such as $\hat{v}=\frac{v}{\parallel v\parallel }$ to make comparison easier. Normalization changes the length but not the direction.

Worked example: a complete 2×2 eigenvalue calculation.

Consider the matrix $A=\left(\begin{array}{cc}3& 1\\ 4& 2\end{array}\right)$. This is a good teaching example because it has two distinct real eigenvalues, so you can see the full workflow without needing complex arithmetic. If you use the calculator above, you would enter $a_{11}=3$, $a_{12}=1$, $a_{21}=4$, and $a_{22}=2$.

Step 1: compute the trace and determinant. The trace is $\mathrm{tr}\left(A\right)=3+2=5$. The determinant is $\det \left(A\right)=3\cdot 2-1\cdot 4=2$. So the characteristic polynomial will be ${\lambda }^2-5\lambda +2$.

Step 2: solve for the eigenvalues. Set the polynomial equal to zero: ${\lambda }^2-5\lambda +2=0$. The quadratic formula gives $\lambda =\frac{5\pm \sqrt{5^2-4\cdot 2}}{2}=\frac{5\pm \sqrt{17}}{2}$. Therefore ${\lambda }_1=\frac{5+\sqrt{17}}{2}$ and ${\lambda }_2=\frac{5-\sqrt{17}}{2}$.

Step 3: find an eigenvector for the first eigenvalue. Now solve $(A-{\lambda }_{1}I)v=0$. Using the first row gives $(3-{\lambda }_1)x+y=0$, so $y=({\lambda }_1-3)x$. If you choose $x=1$, one valid eigenvector is $v_1=\left(\begin{array}{c}1\\ {\lambda }_1-3\end{array}\right)$.

Step 4: find an eigenvector for the second eigenvalue. Repeat the same step with ${\lambda }_2$. Solving $(A-{\lambda }_{2}I)v=0$ leads to $y=({\lambda }_2-3)x$. Setting $x=1$ gives $v_2=\left(\begin{array}{c}1\\ {\lambda }_2-3\end{array}\right)$.

The interpretation is the important part. Vectors aligned with $v_1$ are stretched by ${\lambda }_1$, while vectors aligned with $v_2$ are stretched by ${\lambda }_2$. In repeated multiplication, the direction with larger magnitude usually dominates the long term picture.

How to interpret the calculator output.

When the calculator returns an answer, try reading the output as a story rather than a checklist. Start with the eigenvalues. Their signs and magnitudes tell you whether the special directions grow, shrink, or flip. Then look at the eigenvectors. Those are the actual directions in the plane that survive the transformation. Finally, compare the trace and determinant with the eigenvalues to build intuition and to catch data-entry mistakes. If the trace is wildly different from the sum of the eigenvalues you expected, or the determinant sign seems wrong, it is worth checking the matrix entries again.

This is also where repeated eigenvalues deserve care. If ${\lambda }_1={\lambda }_2$, the matrix may still have two independent eigenvectors, but it may also have only one eigendirection. In that defective case, the calculator can show the same real direction for both roots or give limited information about distinct real eigenvectors. That is not a bug in the calculator. It reflects the underlying algebra.

Why people use eigenvalues in practice.

Many real problems can be simplified by understanding the special directions of a matrix. In an iterative model such as $x_{n+1}=A{x}_n$, the largest magnitude eigenvalue often predicts the dominant long term trend. If $\left|{\lambda }_1\right|>\left|{\lambda }_2\right|$, repeated powers $A^k$ tend to emphasize the direction of $v_1$. That is the intuition behind the power method and many growth or decay analyses.

In a matrix that can be diagonalized, one writes $A=PD{P}^{-1}$, where the columns of $P$ are eigenvectors and $D$ holds the eigenvalues on the diagonal. That decomposition makes powers of the matrix easier to understand because $A^k=P{D}^k{P}^{-1}$. Even in a small 2×2 setting, this explains why some directions dominate after many iterations while others fade away.

Engineers meet the same ideas in vibration problems, control, and linearized systems. Data students meet them in principal directions and repeated transformations. The 2×2 case is not just a toy problem. It is the cleanest place to learn the core idea before moving to larger matrices and more advanced numerical methods.

Quick formula recap. If you want a compact summary after using the tool, these are the relationships worth remembering: $Av=\lambda v$, $\det (A-\lambda I)=0$, $\mathrm{tr}\left(A\right)={\lambda }_1+{\lambda }_2$, $\det \left(A\right)={\lambda }_1{\lambda }_2$, $A{v}_1={\lambda }_1{v}_1$, $A{v}_2={\lambda }_2{v}_2$, $x_{n+1}=A{x}_n$, and, when one eigenvalue dominates, $A^{k}x\approx c{{\lambda }_1}^k{v}_1$. Those formulas are the bridge between the numeric output and the geometric picture.

Common questions.

Can a 2×2 matrix have complex eigenvalues? Yes. If , the roots of the characteristic polynomial are complex. In that case the eigenvalues appear as a conjugate pair, and there is no real eigenvector line in the plane. The calculator reports the complex values numerically and makes it clear that the real-eigenvector interpretation no longer applies.

Are eigenvectors unique? No. If $v$ is an eigenvector, then any nonzero multiple $cv$ with $c\ne 0$ is also an eigenvector for the same eigenvalue. That is why the calculator shows a normalized vector. It is choosing one convenient representative of an entire line of valid answers.

What happens when the matrix has a repeated eigenvalue? If the characteristic polynomial has a double root, then ${\lambda }_1={\lambda }_2$. Sometimes there are still two independent eigenvectors, and sometimes there is only one eigendirection. The second case is called defective. In a 2×2 defective matrix, the repeated eigenvalue does not give two different invariant lines.

How does this connect to stability analysis? For a discrete process such as $x_{n+1}=A{x}_n$, the magnitudes of the eigenvalues help determine whether trajectories grow or decay. If both magnitudes are below 1, solutions tend to shrink. If one magnitude is above 1, that eigendirection can dominate the long term behavior. The calculator gives the raw values; your interpretation comes from the model that produced the matrix.