Diagonalization of a Matrix | 矩阵的对角化

📚 Diagonalization of a Matrix | 矩阵的对角化

Diagonalization is a central concept in linear algebra, enabling us to express a square matrix as a product of three matrices: an invertible matrix, a diagonal matrix, and the inverse of the invertible matrix. This decomposition simplifies many computations, such as raising a matrix to a power or solving systems of linear differential equations.

对角化是线性代数中的核心概念，它使我们能够将一个方阵表示为三个矩阵的乘积：一个可逆矩阵、一个对角矩阵以及该可逆矩阵的逆矩阵。这一分解能够简化许多运算，例如计算矩阵的幂或者求解线性微分方程组。

In this article, we will explore the conditions for diagonalization, the roles of eigenvalues and eigenvectors, and the steps required to perform the process. We will also examine cases where diagonalization is not possible and discuss applications that highlight its importance.

在本文中，我们将探讨对角化的条件、特征值与特征向量的作用，以及执行该过程的步骤。我们还将研究无法对角化的情况，并讨论彰显其重要性的应用。

---

1. What Is Diagonalization? | 什么是对角化？

contains the eigenvalues of A on its main diagonal, and the columns of P are the corresponding eigenvectors.

对于一个 n×n 的方阵 A，如果存在一个可逆矩阵 P 和一个对角矩阵 D，使得 A = P D P⁻¹，则称 A 是可对角化的。矩阵 D 的主对角线上是 A 的特征值，而 P 的各列是对应的特征向量。

nt, and eigenvalues. The formula A = P D P⁻¹ is the key to unlocking simpler calculations.

这种表示称为相似变换，它保留了 A 的许多性质，如迹、行列式和特征值。公式 A = P D P⁻¹ 是释出简化计算的关键。

When a matrix is diagonalizable, operations like A^k become straightforward: A^k = P D^k P⁻¹, where D^k is simply the diagonal matrix with each eigenvalue raised to the power k.

当矩阵可对角化时，像 A^k 这样的运算就变得简单：A^k = P D^k P⁻¹，其中 D^k 只是每个特征值取 k 次幂构成的对角矩阵。

---

2. Eigenvalues and Eigenvectors | 特征值与特征向量

a non-zero vector v such that Av = λv. The vector v is the eigenvector associated with λ.

对角化的基础在于特征值与特征向量。若存在一个非零向量 v，使得 Av = λv，则标量 λ 称为 A 的特征值。向量 v 就是对应于 λ 的特征向量。

Geometrically, an eigenvector is a direction that is left unchanged (or simply scaled) by the linear transformation represented by A. The eigenvalue indicates the factor by which the eigenvector is stretched or compressed.

从几何角度看，特征向量是在 A 所表示的线性变换下保持方向不变（或仅被缩放）的方向。特征值则表示该特征向量被拉伸或压缩的倍数。

pically expect up to n linearly independent eigenvectors, but this is not always the case.

求取特征值和特征向量是实现对角化的第一步。对于一个 n×n 矩阵，我们通常期望最多有 n 个线性无关的特征向量，但这并非总是成立。

---

3. The Diagonalization Condition | 对角化条件

A matrix A is diagonalizable if and only if it has n linearly independent eigenvectors, where n is the size of the matrix. This is equivalent to saying that the eigenvectors can form a basis for the vector space Rⁿ (or Cⁿ).

当且仅当一个矩阵 A 具有 n 个线性无关的特征向量（n 为矩阵的阶数）时，它才是可对角化的。这等价于特征向量可以构成向量空间 Rⁿ（或 Cⁿ）的一组基。

The condition can also be expressed in terms of algebraic and geometric multiplicities. For each eigenvalue λ, the algebraic multiplicity is its multiplicity as a root of the characteristic polynomial; the geometric multiplicity is the dimension of its eigenspace (the null space of A - λI). A is diagonalizable if and only if for each eigenvalue, the geometric multiplicity equals the algebraic multiplicity.

该条件也可用代数重数与几何重数来表达。对每个特征值 λ，代数重数是它作为特征多项式根的重数；几何重数是其特征空间（即 A - λI 的零空间）的维数。当且仅当对每个特征值，几何重数等于代数重数时，A 才是可对角化的。

If a matrix has distinct eigenvalues, it is automatically diagonalizable because the eigenvectors are guaranteed to be linearly independent. Repeated eigenvalues require careful checking of the eigenspace dimensions.

如果一个矩阵具有互不相同的特征值，则它自动可对角化，因为特征向量必定线性无关。若有重特征值，则需要仔细检查特征空间的维数。

---

4. Finding Eigenvalues: The Characteristic Equation | 求特征值：特征方程

To find the eigenvalues, we solve the characteristic equation:

det(A - λI) = 0

det denotes the determinant. The expression det(A - λI) is a polynomial of degree n in λ, called the characteristic polynomial. Its roots are the eigenvalues of A.

此处 I 为 n×n 单位矩阵，det 表示行列式。表达式 det(A - λI) 是一个关于 λ 的 n 次多项式，称为特征多项式。它的根就是 A 的特征值。

ynomial is λ² - (a+d)λ + (ad-bc) = 0, where a+d is the trace and ad-bc is the determinant.

举例，对于 2×2 矩阵 A = [[a, b], [c, d]]，特征多项式为 λ² - (a+d)λ + (ad-bc) = 0，其中 a+d 是迹，ad-bc 是行列式。

Eigenvalues can be real or complex. Even if a matrix has real entries, it may possess complex eigenvalues, which always occur in conjugate pairs. Diagonalization over the real numbers requires all eigenvalues to be real, but we can always diagonalize over the complex numbers if the eigenvector condition is met.

特征值可以是实数或复数。即使矩阵元素均为实数，也可能出现复数特征值，并且复特征值总是共轭成对出现。在实数域上对角化要求所有特征值均为实数，但只要特征向量条件满足，我们总是可以在复数域上实现对角化。

---

5. Finding Eigenvectors: Null Spaces | 求特征向量：零空间

0. The set of all eigenvectors associated with λ, together with the zero vector, forms the eigenspace E_λ = null(A - λI).

一旦求得特征值 λ，其特征向量就是齐次方程组 (A - λI)v = 0 的非零解。与 λ 相关的所有特征向量连同零向量一起构成特征空间 E_λ = null(A - λI)。

To find a basis for the eigenspace, we row-reduce the matrix (A - λI) and express the solution in parametric vector form. The number of free variables gives the geometric multiplicity.

为了找到特征空间的基，我们对矩阵 (A - λI) 进行行化简，并用参数向量形式表示解。自由变量的个数就给出了几何重数。

For diagonalization, we need to collect one eigenvector per linearly independent direction. If an eigenvalue has geometric multiplicity greater than 1, we can choose multiple basis vectors for that eigenspace.

对于对角化，我们需要为每个线性无关的方向收集一个特征向量。若某特征值的几何重数大于 1，我们可以为该特征空间选取多个基向量。

---

6. Constructing the Diagonalization: P and D Matrices | 构造对角化：P 与 D 矩阵

Suppose we have found n linearly independent eigenvectors v₁, v₂, ..., v_n with corresponding eigenvalues λ₁, λ₂, ..., λ_n. Then we form:

P = [v₁ v₂ ... v_n],   D = diag(λ₁, λ₂, ..., λ_n)

The columns of P are precisely the eigenvectors, and the diagonal entries of D are the eigenvalues in the same order. Because the columns are independent, P is invertible.

P 的各列正是特征向量，D 的对角元素则是按相同顺序排列的特征值。由于各列线性无关，P 是可逆的。

= P D P⁻¹. We can verify this by confirming that A P = P D: the j-th column of A P is A v_j, and the j-th column of P D is λ_j v_j, which are equal since A v_j = λ_j v_j.

接着，对角化由 A = P D P⁻¹ 给出。我们可以通过验证 A P = P D 来确认：A P 的第 j 列是 A v_j，P D 的第 j 列是 λ_j v_j，由 A v_j = λ_j v_j 可知二者相等。

---

广告 / Sponsored

7. Step-by-Step Example: Diagonalizing a 2×2 Matrix | 逐步示例：对 2×2 矩阵对角化

Let us diagonalize the matrix A = [[4, 2], [1, 3]]. Follow these steps:

让我们对角化矩阵 A = [[4, 2], [1, 3]]。请按下列步骤操作：

• Step 1: Characteristic equation. det(A - λI) = |4-λ, 2; 1, 3-λ| = (4-λ)(3-λ) - 2 = λ² - 7λ + 10 = 0. The eigenvalues are λ₁ = 2 and λ₂ = 5.
• 步骤1：特征方程。 det(A - λI) = |4-λ, 2; 1, 3-λ| = (4-λ)(3-λ) - 2 = λ² - 7λ + 10 = 0。特征值为 λ₁ = 2 和 λ₂ = 5。

• Step 2: Eigenvector for λ=2. (A - 2I)v = 0: [[2, 2], [1, 1]] → x + y = 0. Choosing x = 1 gives y = -1, so v₁ = (1, -1)^ᵀ.
• 步骤2：对应 λ=2 的特征向量。 解 (A - 2I)v = 0：[[2, 2], [1, 1]] → x + y = 0。取 x = 1 则 y = -1，得 v₁ = (1, -1)^ᵀ。

• Step 3: Eigenvector for λ=5. (A - 5I)v = [[-1, 2], [1, -2]] → -x + 2y = 0. Choose y = 1, then x = 2, giving v₂ = (2, 1)^ᵀ.
• 步骤3：对应 λ=5 的特征向量。 (A - 5I)v = [[-1, 2], [1, -2]] → -x + 2y = 0。选 y = 1，则 x = 2，得 v₂ = (2, 1)^ᵀ。

• Step 4: Form P and D. P = [v₁ v₂] = [[1, 2], [-1, 1]], D = diag(2, 5).
• 步骤4：构造 P 与 D。 P = [v₁ v₂] = [[1, 2], [-1, 1]]，D = diag(2, 5)。

• Step 5: Verify. 1, 1]] (since det(P)=3). Then P D P⁻¹ = [[4,2],[1,3]] = A, confirming the diagonalization.
• 步骤5：验证。 计算 P⁻¹ = (1/3)[[1, -2], [1, 1]]（因为 det(P)=3）。那么 P D P⁻¹ = [[4,2],[1,3]] = A，确认对角化成功。

This example illustrates the standard process. Because the eigenvalues were distinct, diagonalization was guaranteed.

这个例子展示了标准过程。由于特征值互不相同，对角化是必然的。

---

8. The Diagonalization Formula and Its Verification | 对角化公式及其验证

citly, we can check that AP = PD. This equality holds because for each column index j, A v_j = λ_j v_j.

一旦得到 P 和 D，便可直接使用关系式 A = PDP⁻¹。若要避免显式计算 P⁻¹ 来进行验证，我们可以检验 AP = PD 是否成立。由于对每个列索引 j 均有 A v_j = λ_j v_j，该等式成立。

r technology for larger matrices) will confirm the original matrix. This step is often omitted, but it reinforces understanding of the mechanics.

如果对角化正确，那么通过 P 计算出 P⁻¹（例如 2×2 情形使用伴随矩阵法，或对更大矩阵借助技术工具）将会还原出原矩阵。这一步常被省略，但它能强化对机制的理解。

It is important to match the ordering of eigenvalues in D with the columns of P. Swapping two eigenvectors and their corresponding eigenvalues yields another valid diagonalization.

务必使 D 中特征值的顺序与 P 的列相匹配。交换两个特征向量及其对应的特征值会产生另一种有效的对角化。

---

9. Non-Diagonalizable Matrices: The Defective Case | 不可对角化矩阵：亏损矩阵

1]]. Its characteristic polynomial is (1-λ)², giving λ = 1 with algebraic multiplicity 2.

若一个矩阵未能拥有 n 个线性无关的特征向量，则称其为亏损矩阵。经典的例子是 A = [[1, 1], [0, 1]]。其特征多项式为 (1-λ)²，得到 λ = 1，代数重数为 2。

onent to be zero. Thus the eigenspace consists of all vectors (x, 0) and is only 1-dimensional (geometric multiplicity 1). Since 1 < 2, A is not diagonalizable.

求解 (A - I)v = 0 得到 [[0, 1], [0, 0]]v = 0，这迫使第二个分量必须为零。因此特征空间由所有形如 (x, 0) 的向量组成，仅为 1 维（几何重数为 1）。由于 1 < 2，A 不可对角化。

Defective matrices can still be triangularized (e.g., via Jordan normal form), but they lack a full set of eigenvectors. Recognizing this early saves effort when applying diagonalization techniques.

亏损矩阵仍然可以被三角化（例如通过若尔当标准形），但它们缺少一组完整的特征向量。在运用对角化技巧时，及早认识到这一点可以避免徒劳。

---

10. Applications: Powers of Matrices and Differential Equations | 应用：矩阵的幂与微分方程

One of the most practical uses of diagonalization is computing A^k for large k. Instead of multiplying A by itself repeatedly, we use A^k = P D^k P⁻¹. Since D is diagonal, D^k simply contains λ_i^k on the diagonal.

对角化最实际的应用之一是计算大数次幂的 A^k。我们无需重复自乘 A，而是使用 A^k = P D^k P⁻¹。因为 D 是对角矩阵，D^k 只需在对角线上置入 λ_i^k 即可。

This approach is essential in discrete dynamical systems, such as Markov chains or population models, where the state vector after k steps is given by A^k x₀. Diagonalization allows explicit closed-form solutions.

这一方法在离散动力系统中至关重要，例如马尔可夫链或种群模型，其中经过 k 步后的状态向量由 A^k x₀ 给出。对角化使得我们可以得到显式的闭式解。

by substituting x = P y, leading to dy/dt = D y, which consists of independent equations y_i' = λ_i y_i. The solution is then y_i(t) = c_i e^{λ_i t}, and x(t) = P y(t).

对于微分方程组 dx/dt = A x，可通过代换 x = P y 实现解耦，从而得到 dy/dt = D y，它由相互独立的方程 y_i' = λ_i y_i 组成。于是解为 y_i(t) = c_i e^{λ_i t}，而 x(t) = P y(t)。

These applications highlight why diagonalization is not merely a theoretical exercise but a powerful tool in applied mathematics.

这些应用凸显出对角化不仅是理论推演，更是应用数学中的强大工具。

---

11. Complex Eigenvalues and Diagonalization over ℂ | 复特征值与在复数域上的对角化

When a real matrix has complex eigenvalues, diagonalization is generally performed over the complex numbers ℂ. For instance, a rotation matrix in R² has eigenvalues e^±iθ and corresponding complex eigenvectors. Over ℂ, it becomes diagonalizable.

当实矩阵具有复特征值时，对角化通常在复数域 ℂ 上进行。例如，R² 中的旋转矩阵就具有特征值 e^±iθ 和相应的复特征向量。在 ℂ 上，它便是可对角化的。

The algebraic process remains the same: find eigenvalues from the characteristic polynomial, then determine eigenvectors using complex arithmetic. The resulting P and D will contain complex entries.

代数过程保持不变：从特征多项式求出特征值，然后利用复数运算确定特征向量。最终得到的 P 和 D 将包含复数元素。

In many IB contexts, diagonalization is restricted to real eigenvalues, but it is helpful to be aware that the fundamental definitions extend naturally to the complex case, ensuring that matrices with simple but complex eigenvalues are still diagonalizable.

在许多 IB 课程情境中，对角化仅限于实特征值，但了解基本定义可以自然地延拓到复数情形是有益的，这保证了那些具有简单但复特征值的矩阵仍然是可对角化的。

---

12. Summary and Key Takeaways | 总结与关键要点

re ingredients are eigenvalues and eigenvectors. A matrix is diagonalizable if and only if it possesses a full set of n linearly independent eigenvectors.

对角化通过相似变换 A = PDP⁻¹ 将矩阵 A 化为对角形式。其核心要素是特征值与特征向量。当且仅当矩阵具有完整的 n 个线性无关的特征向量时，它才是可对角化的。

the eigenvectors and D from the eigenvalues; (4) verify using AP = PD or by computing P⁻¹.

步骤为：(1) 通过 det(A - λI) = 0 计算特征值；(2) 对每个特征值求特征空间的基；(3) 用特征向量组成 P，用特征值组成 D；(4) 通过 AP = PD 或计算 P⁻¹ 进行验证。

Key pitfalls include insufficient independent eigenvectors (defective matrices) and mismatching eigenvalues with the wrong vectors. Mastering diagonalization prepares students for advanced topics such as principal component analysis, quantum mechanics, and stability analysis.

需要留意的关键陷阱包括特征向量不足（亏损矩阵）以及特征值与错误向量匹配不当。掌握对角化将为学生深入学习主成分分析、量子力学和稳定性分析等高级课题做好准备。

Published by TutorHao | Mathematics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）