📚 Eigenvectors | 特征向量
Eigenvectors sit at the heart of linear algebra, capturing the 'direction' that remains unchanged when a linear transformation is applied. For any square matrix A, a non‑ d an eigenvector of A, and the scalar λ is the corresponding eigenvalue. Understanding eigenvectors unlocks the ability to diagonalise matrices, solve systems of differential equations, perform principal component analysis, and much more. In the International Baccalaureate (IB) Mathematics curriculum, eigenvectors and eigenvalues appear in the Analysis and Approaches Higher Level course, where they bridge pure algebraic manipulation with geometric insight. This article explores the definition, computation, properties, geometric meaning, and applications of eigenvectors in a clear, step‑by‑step manner, helping you master this fundamental concept.
特征向量是线性代数的核心概念之一，它表示在线性变换下“方向”保持不变的向量。对于任意方阵A，若存在非零向量v满足Av = λv，则称v为A的特征向量，标量λ为对应的特征值。理解特征向量使我们能够对角化矩阵、求解微分方程组、进行主成分分析等。在国际文凭（IB）数学课程中，特征向量与特征值出现在分析与方法的高级水平中，连接着纯代数运算与几何直观。本文将从定义、计算、性质、几何意义和应用等方面，逐步深入地解析特征向量，帮助你扎实掌握这一基础概念。
1. The Core Definition | 核心定义
An eigenvector of an n × n matrix A is a non‑zero column vector v ∈ ℝⁿ (or ℂ change (unless λ is negative, in which case it reverses, but still lies on the same line). The scalar λ is called the eigenvalue associated with v. Crucially, if v is an eigenvector, any non‑ se A(k v) = k A v = k λ v = λ (k v). Thus eigenvectors are defined up to a scalar multiple – they represent an eigen‑direction rather than a single vector.
对于n × n矩阵A，若存在非零列向量v ∈ ℝⁿ（或ℂⁿ）和标量λ，使得A v = λ v，则称v为A的特征向量。这个等式表明，A作用于v时，输出仅是v的缩放版本；v的方向保持不变（若λ为负则反向，但仍在同一直线上）。标量λ称为对应于v的特征值。重要的是，如果v是特征向量，那么任何非零倍数k v也是同一特征值下的特征向量，因为A(k v) = k A v = k λ v = λ (k v)。因此，特征向量在标量倍数意义下是等价的——它们代表一个特征方向，而非单一的向量。
2. Finding Eigenvectors via the Characteristic Equation | 通过特征方程求特征向量
es. Rearranging A v = λ v gives (A − λ I) v = 0, where I is the identity matrix. For a non‑ s is the characteristic equation, a polynomial in λ. Solving it yields eigenvalues λ₁, λ₂, … f all eigenvectors associated with λᵢ, together with the zero vector, forms the eigenspace Eᵢ.
为了求特征向量，我们需要先确定特征值。将A v = λ v移项得(A − λ I) v = 0，其中I为单位矩阵。要使存在非零解v，矩阵(A − λ I)必须是奇异的，因此其行列式必须为零：det(A − λ I) = 0。这就是特征方程，是关于λ的多项式。解之得到特征值λ₁, λ₂, …, λₙ。对于每个特征值λᵢ，将其代回(A − λᵢ I) v = 0，解齐次线性方程组求出对应的特征向量。与λᵢ关联的所有特征向量加上零向量构成的集合称为特征空间Eᵢ。
3. Worked Example: 2 × 2 Matrix | 实例：2 × 2 矩阵
= [ [1, 1], [1, 1] ] v = 0. This gives x + y = 0, so y = −x. An eigenvector is v = (1, −1)ᵀ [ [−1, 1], [1, −1] ] v = 0, which gives −x + y = 0, so y = x. An eigenvector is v = (1, 1)ᵀ. Notice these two eigenvectors are orthogonal, a property linked to the symmetry of A.
考虑矩阵A = [ [2, 1], [1, 2] ]。构造A − λ I = [ [2−λ, 1], [1, 2−λ] ]。行列式为(2−λ)² − 1 = λ² − 4λ + 3 = (λ − 1)(λ − 3)。因此特征值为λ = 1和λ = 3。对于λ = 1，解(A − I) v = [ [1, 1], [1, 1] ] v = 0，得x + y = 0，即y = −x。特征向量可取v = (1, −1)ᵀ。对于λ = 3，解(A − 3I) v = [ [−1, 1], [1, −1] ] v = 0，得−x + y = 0，即y = x。特征向量可取v = (1, 1)ᵀ。注意到这两个特征向量相互正交，这与A的对称性有关。
4. Geometric Picture: Stretching, Compressing and Reflecting | 几何图像：拉伸、压缩与反射
at is preserved by the transformation represented by A. If λ > 1, vectors along that eigen‑ transformation can reveal how eigenvectors define the principal axes of the transformation. For instance, the matrix above maps the square to a sheared shape, but along the lines y = x and y = −x, points simply scale.
从几何上看，特征向量指示了在矩阵A所代表的线性变换下保持不变的方向。若λ > 1，沿该特征方向的向量被拉伸；若0 < λ < 1，则被压缩；若λ = 1，则保持不变；若λ = 0，则被压缩为零向量（零空间）；若λ为负，则方向反转。在二维空间中，矩阵通常有两个互异的特征向量。通过观察单位正方形在变换下的形状变化，可以看出特征向量如何定义变换的主轴。例如，上述矩阵将正方形映射为一个倾斜的形状，但沿着直线y = x和y = −x，点仅发生了缩放。
5. Algebraic Multiplicity vs Geometric Multiplicity | 代数重数与几何重数
ve geometric multiplicity equal to its algebraic multiplicity. If geometric multiplicity is less, the matrix is defective and cannot be diagonalised, though it can be put into Jordan normal form. For example, A = [ [1, 1], [0, 1] ] has eigenvalue 1 with algebraic multiplicity 2, but solving (A − I)v = 0 gives only one linearly independent eigenvector (1, 0)ᵀ, so geometric multiplicity is 1. This matrix is not diagonalisable.
当特征值作为特征多项式的重根出现时，我们需要区分其代数重数（它作为根出现的次数）和几何重数（其对应特征空间的维数）。一个矩阵可对角化的条件是：每个特征值的几何重数等于其代数重数。若几何重数小于代数重数，则该矩阵是亏损的，无法对角化，但可化为若尔当标准形。例如，A = [ [1, 1], [0, 1] ]的特征值为1，代数重数为2，但解(A − I)v = 0仅得到一个线性无关的特征向量(1, 0)ᵀ，因此几何重数为1。该矩阵不可对角化。
6. Eigenvectors of Symmetric Matrices | 对称矩阵的特征向量
Real symmetric matrices (A = Aᵀ es are orthogonal. Moreover, a real symmetric matrix is always orthogonally diagonalisable – there exists an orthogonal matrix P (P⁻¹ = Pᵀ) such that Pᵀ f inertia. In IB, you may be asked to verify orthogonality of computed eigenvectors by taking their dot product, which should be zero for distinct eigenvalues.
实对称矩阵（A = Aᵀ）具有优异的性质：其特征值均为实数，且对应不同特征值的特征向量彼此正交。此外，实对称矩阵总是可正交对角化——存在正交矩阵P（P⁻¹ = Pᵀ），使得Pᵀ A P为对角阵。这是谱定理的基础，支撑着二次型、主轴旋转和转动惯量的物理解释。在IB课程中，你可能需要验证计算所得特征向量的正交性：对属于不同特征值的向量计算点积，结果应为零。
7. Diagonalisation and Powers of a Matrix | 对角化与矩阵的幂
If an n × n matrix A has n linearly independent eigenvectors v₁, v₂, …, vₙ, with corresponding eigenvalues λ₁, λ₂, … the matrix P whose columns are the eigenvectors, and the diagonal matrix D = diag(λ₁, λ₂, … nsely useful for computing powers of A: Aᵏ = P Dᵏ P⁻¹, because Dᵏ is simply diag(λ₁ᵏ, λ₂ᵏ, … quence can be expressed using powers of a 2 × 2 matrix, diagonalised via its eigenvectors.
如果一个n × n矩阵A有n个线性无关的特征向量v₁, v₂, …, vₙ，对应特征值为λ₁, λ₂, …, λₙ，则可将特征向量组成矩阵P的列，并构建对角矩阵D = diag(λ₁, λ₂, …, λₙ)。于是A = P D P⁻¹。这种分解在计算矩阵的幂时极其有用：Aᵏ = P Dᵏ P⁻¹，因为Dᵏ就是diag(λ₁ᵏ, λ₂ᵏ, …, λₙᵏ)。它还简化了递推关系和离散动力系统的求解。例如，斐波那契数列可用一个2×2矩阵的幂表示，并通过其特征向量进行对角化求解。
8. Eigenvectors in Differential Equations | 特征向量在微分方程中的应用
ly solved using eigenvectors. If v is an eigenvector of A with eigenvalue λ, then x(t) = eλt v is a solution. The general solution is a linear combination of such terms. This method transforms a coupled system into uncoupled equations along eigen‑directions. In IB, you encounter this when modelling population dynamics, mixing problems, or coupled oscillations. Understanding eigenvectors turns a seemingly complex system of ODEs into a manageable superposition of exponential modes.
线性微分方程组，例如dx/dt = A x，可自然地利用特征向量求解。如果v是A的特征向量，对应特征值为λ，则x(t) = eλt v是一个解。通解就是这类项的线性组合。该方法将耦合系统沿着特征方向解耦为独立的方程。在IB课程中，你会遇到用特征向量建模人口动力学、混合问题或耦合振荡的情形。理解特征向量能将表面上复杂的一阶微分方程组化为可处理的指数模态叠加。
9. Complex Eigenvalues and Rotation | 复特征值与旋转
he rotation matrix [ [cos θ, −sin θ], [sin θ, cos θ] ] has eigenvalues cos θ ± i sin θ = e±iθ anges direction. Complex eigenvectors, though less visual, are essential for understanding spiral sinks and sources in phase portraits of dynamical systems.
并非所有矩阵都有实特征值。一个2×2实矩阵可以有一对共轭复特征值a ± b i。此时特征向量也是复的。从几何上讲，这对应于包含旋转变换。例如，旋转矩阵[ [cos θ, −sin θ], [sin θ, cos θ] ]的特征值为cos θ ± i sin θ = e±iθ。除了θ = 0或π的情况，该矩阵没有实特征向量，因为每个向量都改变了方向。复特征向量虽然不那么直观，但对于动力系统相图中螺旋汇和螺旋源的理解至关重要。
10. Invariant Lines and the Determinant‑Trace Connection | 不变直线与行列式‑迹的关系
if det(A) = 0, at least one eigenvalue is zero, meaning there is a whole line of points mapped to the origin – ind invariant lines of a transformation, which are precisely the lines along eigenvectors.
在二维情况下，特征向量对应变换下过原点的不变直线。特征值由矩阵A的迹（对角线元素之和）和行列式决定：λ₁ + λ₂ = tr(A)，λ₁ λ₂ = det(A)。这为检验计算出的特征值提供了快速方法。此外，若det(A) = 0，至少有一个特征值为零，这意味着存在整条直线上的点被映射到原点——相应的特征向量张成零空间。IB试题常要求找出变换的不变直线，这些直线正是沿着特征向量的直线。
11. Principal Component Analysis (PCA) – A Glimpse of Data Science | 主成分分析——数据科学一瞥
Eigenvectors are not confined to pure mathematics; they are the engine of Principal Component Analysis, a statistical technique for dimensionality reduction. Given a data covariance matrix, its eigenvectors represent the directions of maximum variance (principal components), and the corresponding eigenvalues indicate the magnitude of variance along those directions. By projecting data onto the first few eigenvectors, we reduce dimensionality while preserving information. This connection shows why eigenvectors are fundamental in machine learning, image compression, and quantitative finance, making the IB topic surprisingly forward‑looking.
特征向量不仅限于纯数学领域，它还是主成分分析（一种降维统计技术）的引擎。对于数据协方差矩阵，其特征向量代表最大方差的方向（主成分），对应的特征值则表示沿这些方向的方差大小。通过将数据投影到前几个特征向量上，我们在保留信息的同时降低了维度。这种联系解释了为何特征向量在机器学习、图像压缩和量化金融中如此基础，也使IB的这一课题颇具前瞻性。
12. Common Pitfalls and Tips for IB Exams | IB考试的常见陷阱与应考技巧
Students often confuse eigenvectors with eigenvalues, forget to state 'non‑zero vector' lving (A − λ I)v = 0, always present the solution as a general vector, e.g., v = t (1, −1)ᵀ, t ≠ 0. Watch out for arithmetic errors when computing determinants; double‑ ors geometrically: sketch the invariant lines. Lastly, in problems that ask for Aⁿ, use diagonalisation if possible, but only after confirming the matrix is diagonalisable (i.e., enough independent eigenvectors).
学生常犯的错误包括混淆特征向量与特征值、忘记说明“非零向量”、或将特征向量错误地写成矩阵形式。在解(A − λ I)v = 0时，一定要将解写成通解形式，如v = t (1, −1)ᵀ, t ≠ 0。注意计算行列式时的代数错误，可利用迹与行列式的关系进行快速检验。若矩阵对称，用点积验证正交性。练习从几何角度解释特征向量：画出不变直线。最后，在涉及Aⁿ的问题中，若可能则使用对角化，但先要确认矩阵可对角化（即有足够多的线性无关特征向量）。
Published by TutorHao | Mathematics Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图