Eigenvectors: Solving Methods and Geometric Interpretation | 特征向量的求解方法与几何意义

📚 Eigenvectors: Solving Methods and Geometric Interpretation | 特征向量的求解方法与几何意义

Eigenvectors and eigenvalues are among the most powerful concepts in linear algebra, forming the backbone of matrix diagonalization, differential equations, and data science algorithms such as Principal Component Analysis (PCA). In this article, we will explore the systematic methods for finding eigenvectors and uncover their profound geometric meaning in linear transformations.

特征向量与特征值是线性代数中最重要的概念之一，它们是矩阵对角化、微分方程求解以及主成分分析（PCA）等数据科学算法的基石。本文将系统介绍特征向量的求解方法，并深入揭示其在线性变换中的几何内涵。

---

1. Definition of Eigenvectors and Eigenvalues | 特征向量与特征值的定义

ix. A nonzero vector v is called an eigenvector of A if there exists a scalar λ such that:

Av = λv

Here, λ is called the eigenvalue associated with the eigenvector v. The key property is that applying the transformation A to v simply scales v by a factor of λ, without changing its direction (unless λ is negative, in which case the direction is reversed).

设 A 是一个 n × n 方阵。若非零向量 v 满足存在一个标量 λ，使得 Av = λv，则称 v 为 A 的特征向量，λ 为对应的特征值。其核心性质是：对 v 施加线性变换 A，仅仅是将 v 缩放 λ 倍，方向不发生改变（若 λ 为负数，则方向反转）。

• nonzero. The zero vector trivially satisfies A0 = λ0 but is not considered an eigenvector.
• v 必须是非零向量。零向量虽然满足 A0 = λ0，但通常不将其视为特征向量。
• There are infinitely many eigenvectors for a given eigenvalue, since any scalar multiple of an eigenvector is also an eigenvector.
• 对应同一特征值的特征向量有无数个，因为特征向量的任意非零标量倍仍然是特征向量。

---

2. The Characteristic Equation | 特征方程

or a nonzero v to exist, the matrix (A - λI) must be singular, meaning its determinant must be zero:

det(A - λI) = 0

This equation is called the characteristic equation of A. Expanding det(A - λI) produces a polynomial of degree n in λ, known as the characteristic polynomial, whose roots are exactly the eigenvalues of A.

为了求解特征值，我们将等式 Av = λv 改写为 (A - λI)v = 0，其中 I 为单位矩阵。要使非零向量 v 存在，矩阵 (A - λI) 必须是奇异的，即其行列式为零：det(A - λI) = 0。该方程称为 A 的特征方程。展开行列式 det(A - λI) 会得到一个关于 λ 的 n 次多项式，称为特征多项式，其根恰好就是 A 的所有特征值。

For example, consider A = [2 1; 1 2]. The characteristic equation is:

det([2-λ 1; 1 2-λ]) = (2-λ)² - 1 = λ² - 4λ + 3 = 0

The roots are λ = 3 and λ = 1, giving us two distinct eigenvalues.

例如，考虑矩阵 A = [2 1; 1 2]，其特征方程为 det([2-λ 1; 1 2-λ]) = (2-λ)² - 1 = λ² - 4λ + 3 = 0，解得 λ = 3 和 λ = 1，即两个互不相同的特征值。

---

3. Finding Eigenvalues: Step-by-Step | 求解特征值：分步方法

The process of finding eigenvalues can be summarized in three clear steps:

求解特征值的过程可以归纳为三个清晰的步骤：

• Step 1: Write down the matrix A and form A - λI by subtracting λ from each main diagonal entry.
  步骤 1：写出矩阵 A，并将主对角线上的每个元素减去 λ，得到 A - λI。
• Step 2: Compute the determinant det(A - λI) and set it equal to zero to obtain the characteristic polynomial.
  步骤 2：计算行列式 det(A - λI)，并令其等于零，得到特征多项式。
• Step 3: Solve the polynomial equation for λ. Each distinct root is an eigenvalue.
  步骤 3：解这个关于 λ 的多项式方程，每个不同的根就是矩阵的一个特征值。

e polynomial is cubic, and the rational root theorem is often helpful for finding integer roots.

对于 2 × 2 矩阵，特征多项式是二次的，可以通过因式分解或二次公式求解。对于 3 × 3 矩阵，特征多项式是三次的，此时有理根定理常对寻找整数根很有帮助。

---

4. Finding Eigenvectors: Step-by-Step | 求解特征向量：分步方法

Once an eigenvalue λ is known, we find the corresponding eigenvectors by solving the homogeneous system:

(A - λI)v = 0

This is done using Gaussian elimination to find the null space of the matrix (A - λI).

在已知特征值 λ 后，我们通过求解齐次线性方程组 (A - λI)v = 0 来找到对应的特征向量，这通常使用高斯消元法来确定矩阵 (A - λI) 的零空间。

Let us continue with A = [2 1; 1 2]. For λ = 3:

A - 3I = [-1 1; 1 -1]

ector of the form [t; t] is an eigenvector. A convenient choice is v₁ = [1; 1]. For λ = 1:

A - 1I = [1 1; 1 1]

Solving x + y = 0 gives y = -x, so v₂ = [1; -1] is an eigenvector.

继续使用矩阵 A = [2 1; 1 2]。当 λ = 3 时，A - 3I = [-1 1; 1 -1]。解方程 (-1)x + y = 0 得 y = x，因此形如 [t; t] 的向量都是特征向量，通常取 v₁ = [1; 1]。当 λ = 1 时，A - 1I = [1 1; 1 1]，解 x + y = 0 得 y = -x，故取 v₂ = [1; -1] 为一个特征向量。

---

5. Geometric Interpretation | 几何意义

Geometrically, a linear transformation A maps vectors in n-dimensional space to new vectors. Most vectors will change both their direction and their length under this transformation. However, eigenvectors are special: their direction remains invariant (up to a sign), and they are simply stretched or compressed by the eigenvalue λ. This is the essence of what makes eigenvectors "fundamental directions" of the transformation.

从几何角度看，线性变换 A 将 n 维空间中的向量映射到新的向量。大多数向量在这个变换下方向和长度都会改变。然而，特征向量是特殊的：它们的方向保持不變（至多差一个正负号），仅仅是长度被特征值 λ 拉伸或压缩。这正是特征向量作为变换"基本方向"的本质所在。

• If λ > 1, the eigenvector is stretched in the same direction.
• 若 λ > 1，特征向量沿相同方向被拉长。
• If 0 < λ < 1, the eigenvector is compressed.
• 若 0 < λ < 1，特征向量被压缩。
• If λ < 0, the eigenvector is stretched and its direction flips by 180°.
• 若 λ < 0，特征向量被缩放且方向翻转 180°。
• genvector lies in the null space of A; the transformation collapses it to the zero vector.
• 若 λ = 0，特征向量位于 A 的零空间中，变换将其映射为零向量。

For a 2D transformation, imagine a rubber sheet being stretched: the eigenvectors point along the directions that remain aligned with their original grid lines, serving as the natural axes of the transformation.

想象一块橡皮薄膜被拉伸的过程：特征向量指向那些变换后仍与其原方向保持对齐的方向，它们是该变换的天然主轴。

---

6. Special Cases: Repeated and Zero Eigenvalues | 特殊情况：重根与零特征值

Not all matrices have distinct eigenvalues. When the characteristic polynomial has repeated roots, we say the eigenvalue has algebraic multiplicity greater than 1. The dimension of the corresponding eigenspace (the set of all eigenvectors associated with that eigenvalue, plus the zero vector) is called the geometric multiplicity.

并非所有矩阵都有互不相同的特征值。当特征多项式出现重根时，我们称该特征值的代数重数大于 1。对应的特征子空间（与该特征值相关的全部特征向量加上零向量的集合）的维数称为几何重数。

ponent must be 0. Thus, eigenvectors are of the form [t; 0], and the eigenspace has dimension 1, even though the algebraic multiplicity is 2.

例如，考虑矩阵 B = [3 1; 0 3]，其特征多项式为 (3-λ)² = 0，因此 λ = 3 是重特征值。解 (B - 3I)v = 0 得到 [0 1; 0 0]v = 0，这意味着第二个分量为 0。因此，特征向量形如 [t; 0]，特征子空间维数为 1，尽管代数重数为 2。

itionally, a zero eigenvalue indicates that A is singular (non-invertible), as det(A) = 0.

当几何重数小于代数重数时，矩阵不可对角化。此外，零特征值表明矩阵 A 是奇异的（不可逆），因为此时 det(A) = 0。

---

广告 / Sponsored

7. Diagonalization and Applications | 对角化与应用

endent eigenvectors, it can be diagonalized. We construct the matrix P whose columns are these eigenvectors, and the diagonal matrix D whose diagonal entries are the corresponding eigenvalues. Then:

如果一个 n × n 矩阵 A 有 n 个线性无关的特征向量，则它可以被对角化。构造以这些特征向量为列向量的矩阵 P，以及对角线元素为相应特征值的对角矩阵 D，则成立：

A = PDP⁻¹   ⇒   Aᵏ = PDᵏP⁻¹

This is extremely useful for computing high powers of A, solving systems of linear differential equations, and analyzing the stability of dynamical systems. In IB Mathematics AI-HL and AA-HL, diagonalization frequently appears in exam questions involving Markov chains and recurrence relations.

这一性质对计算矩阵的高次幂、求解线性微分方程组以及分析动力系统的稳定性极为有用。在 IB 数学 AI-HL 和 AA-HL 中，对角化经常出现在涉及马尔可夫链和递推关系的考题中。

-1] and D = [3 0; 0 1]. Then A⁵ = P D⁵ P⁻¹, which simplifies the computation considerably.

例如，对于 A = [2 1; 1 2]，有 P = [1 1; 1 -1] 和 D = [3 0; 0 1]，于是 A⁵ = P D⁵ P⁻¹，从而大大简化了计算。

---

8. IB Exam Tips: Common Pitfalls and Strategies | IB 考试要点：常见错误与应对策略

Through years of marking IB exams, certain mistakes appear repeatedly. Being aware of them can save you valuable marks.

通过多年 IB 阅卷经验，以下是一些反复出现的典型错误。了解它们可以帮助你避免失分。

Common Mistake | 常见错误	Correct Approach | 正确做法
Forgetting that eigenvectors must be nonzero.	Always reject the zero vector when writing eigenvectors.
Writing the characteristic equation incorrectly for 3 × 3 matrices.	Use expansion along a row or column with the most zeros to simplify.
Not normalizing eigenvectors when the question asks for unit eigenvectors.	Divide each eigenvector by its Euclidean norm √x² + y² + z².
Confusing algebraic and geometric multiplicity.	Algebraic = multiplicity as a root; geometric = dimension of eigenspace.

at Av = λv for each eigenpair. This takes only seconds and catches many arithmetic errors.

另一个实用策略是：对于每一对特征值与特征向量，用 Av = λv 进行验算。这只需要几秒钟，却可以捕捉到大量计算错误。

---

9. Worked Example: A 3 × 3 Matrix | 完整例题：3 × 3 矩阵

a full 3 × 3 example to consolidate the method. Find the eigenvalues and eigenvectors of:

A = [4 1 0; 0 4 1; 0 0 3]

gonal entries: λ = 4, λ = 4, and λ = 3. The eigenvalue λ = 4 has algebraic multiplicity 2.

由于 A 是上三角矩阵，其特征值就是主对角线上的元素：λ = 4、λ = 4 和 λ = 3。其中 λ = 4 的代数重数为 2。

For λ = 4, solve (A - 4I)v = 0:

[0 1 0; 0 0 1; 0 0 -1]v = 0

eigenspace has dimension 1, so only one linearly independent eigenvector exists for λ = 4.

对于 λ = 4，解 (A - 4I)v = 0：即 [0 1 0; 0 0 1; 0 0 -1]v = 0。解得 v₂ = 0，v₃ = 0，v₁ 自由。因此一个特征向量是 [1; 0; 0]。特征子空间维数为 1，说明 λ = 4 只有一个线性无关的特征向量。

For λ = 3, solve (A - 3I)v = 0:

[1 1 0; 0 1 1; 0 0 0]v = 0

and v₂ + v₃ = 0. Setting v₃ = 1 gives v₂ = -1 and v₁ = 1, so an eigenvector is [1; -1; 1].

对于 λ = 3，解 (A - 3I)v = 0：即 [1 1 0; 0 1 1; 0 0 0]v = 0。解得 v₁ + v₂ = 0 和 v₂ + v₃ = 0。令 v₃ = 1，得 v₂ = -1，v₁ = 1，故一个特征向量为 [1; -1; 1]。

Notice that A has only two linearly independent eigenvectors in total, so it is not diagonalizable. This illustrates the important connection between the multiplicity of eigenvalues and diagonalizability.

注意，A 总共只有两个线性无关的特征向量，因此它不可对角化。这正好说明特征值的重数与矩阵可对角化之间的重要联系。

---

10. Summary and Final Thoughts | 总结与要点回顾

ing the transformation only stretches them without rotating their direction. The general solving procedure is:

特征向量的核心方程为 Av = λv，其含义是变换仅对向量进行拉伸处理而不改变其方向。一般求解流程为：

1. Compute the characteristic polynomial det(A - λI) = 0.
2. 计算特征多项式 det(A - λI) = 0 并求解所有特征值。
3. For each eigenvalue λ, solve (A - λI)v = 0 to find the eigenspace.
4. 对每个特征值 λ，解齐次方程组 (A - λI)v = 0 以找到特征子空间。
5. Choose basis vectors from the eigenspace as representatives of the eigenvectors.
6. 从特征子空间中选取基向量作为特征向量的代表。

Remember: the geometric multiplicity (dimension of eigenspace) may be larger or smaller than the algebraic multiplicity. Mastering the interplay between eigenvalues, eigenvectors, and their geometric meaning is essential not only for IB exams but also for future studies in physics, engineering, and computer science.

请记住：几何重数（特征子空间的维数）可能大于或小于代数重数。深刻理解特征值、特征向量及其几何意义之间的内在联系，不仅对 IB 考试至关重要，对未来在物理、工程和计算机科学领域的学习同样不可或缺。

Published by TutorHao | Mathematics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）