📚 IB Mathematics: Classification of Linear Equilibrium Points | IB数学：线性平衡点分类详解
In the study of differential equations, one of the most powerful tools is the qualitative analysis of equilibrium points. Rather than solving a system explicitly, we can classify the behaviour of solutions near equilibrium points to understand the long-term dynamics of the system.
在微分方程的研究中，最有力的工具之一是对平衡点进行定性分析。我们不必显式求解系统，而是通过分类平衡点附近的解的行为，来理解系统的长期动力学特征。
This article provides a comprehensive guide to classifying linear equilibrium points, a key topic in IB Mathematics AA HL, particularly in the context of systems of first-order linear differential equations. We will explore the mathematics behind eigenvalues, the stability criteria, and the visual interpretation through phase portraits.
本文为IB数学AA HL中线性平衡点分类这一核心考点提供了全面的指南，特别聚焦于一阶线性微分方程组的背景。我们将深入探讨特征值背后的数学原理、稳定性判据以及通过相图的直观解读。
1. What is an Equilibrium Point? | 什么是平衡点？
An equilibrium point (also called a fixed point or critical point) of a system of differential equations is a point where the derivatives of all state variables are zero. At such a point, the system remains stationary forever if it starts exactly there.
微分方程组的平衡点（也称为不动点或临界点）是系统中所有状态变量的导数均为零的点。在这样一个点上，如果系统从该点出发，它将永远保持静止。
Consider a general two-dimensional linear system:
考虑一个一般的二维线性系统：
dx/dt = ax + by, dy/dt = cx + dy
or in matrix form, dX/dt = A·X, where X = (x, y)ᵀ and A = [a b; c d]. The equilibrium point is found by solving A·X = 0. For a non-singular matrix A, the only equilibrium point is the origin (0, 0).
或用矩阵形式表示为 dX/dt = A·X，其中 X = (x, y)ᵀ，A = [a b; c d]。平衡点通过求解 A·X = 0 得到。对于非奇异矩阵 A，唯一的平衡点就是原点 (0, 0)。
When A is singular, there may be a line or a plane of equilibrium points. In this article, we focus on the case where the origin is an isolated equilibrium point, meaning det(A) ≠ 0.
当 A 为奇异矩阵时，可能存在一条线或一个平面的平衡点。在本文中，我们聚焦于原点为孤立平衡点的情况，即 det(A) ≠ 0。
2. The Role of Eigenvalues | 特征值的作用
The classification of equilibrium points rests entirely on the eigenvalues of the coefficient matrix A. Recall that eigenvalues λ are solutions of the characteristic equation det(A − λI) = 0.
平衡点的分类完全依赖于系数矩阵 A 的特征值。回顾一下，特征值 λ 是特征方程 det(A − λI) = 0 的解。
For a 2×2 matrix, the characteristic equation expands to:
对于 2×2 矩阵，特征方程展开为：
λ² − (tr A)λ + det A = 0
where tr A = a + d is the trace and det A = ad − bc is the determinant of A. The discriminant Δ = (tr A)² − 4det A determines the nature of the eigenvalues:
其中 tr A = a + d 是矩阵的迹，det A = ad − bc 是矩阵的行列式。判别式 Δ = (tr A)² − 4det A 决定了特征值的性质：
-
Δ > 0: two distinct real eigenvalues
Δ > 0：两个不同的实特征值
-
Δ = 0: one repeated real eigenvalue
Δ = 0：一个重实特征值
-
Δ < 0: a pair of complex conjugate eigenvalues
Δ < 0：一对共轭复特征值
The signs and values of these eigenvalues determine the stability and type of the equilibrium point at the origin.
这些特征值的符号和数值决定了原点平衡点的稳定性和类型。
3. Stability Definitions | 稳定性定义
Before we classify equilibrium points, let us establish precise definitions of stability. An equilibrium point X* is said to be stable if solutions starting sufficiently close to X* remain close for all time.
在分类平衡点之前，让我们先建立精确的稳定性定义。如果从距离 X* 足够近的地方出发的解在所有时间内都保持在该点附近，则称平衡点 X* 是稳定的。
An equilibrium point is asymptotically stable if it is stable and, in addition, solutions starting sufficiently close to X* converge to X* as t → ∞. If solutions starting arbitrarily close to X* move away, the equilibrium point is unstable.
如果平衡点 X* 是稳定的，并且从足够接近 X* 处出发的解在 t → ∞ 时收敛到 X*，则该平衡点为渐近稳定。如果从任意接近 X* 处出发的解都会远离，则该平衡点是不稳定的。
In terms of eigenvalues for a linear system, asymptotic stability is equivalent to all eigenvalues having negative real parts. Instability occurs if at least one eigenvalue has a positive real part. If all eigenvalues have non-positive real parts and some are purely imaginary, the equilibrium point is stable but not asymptotically.
对于线性系统，用特征值来表述，渐近稳定等价于所有特征值的实部均为负。如果至少有一个特征值的实部为正，则系统不稳定。如果所有特征值的实部都非正且有些是纯虚数，则平衡点稳定但不是渐近稳定。
4. Classification via Trace and Determinant | 利用迹与行列式进行分类
In many textbooks, the classification of equilibrium points is presented in terms of the trace (p) and determinant (q) of the matrix A. This approach is elegant because it requires no explicit eigenvalue computation.
在许多教科书中，平衡点的分类是根据矩阵 A 的迹 p 和行列式 q 来呈现的。这种方法非常优雅，因为它不需要显式计算特征值。
Recall the relationship between eigenvalues and these quantities:
回顾特征值与这些量之间的关系：
λ₁ + λ₂ = tr A = p, λ₁ × λ₂ = det A = q
The parabola q = p²/4 separates the region into distinct types. The following diagram summarises the classification in the (p, q)-plane.
抛物线 q = p²/4 将区域划分为不同类型。下面的图表总结了 (p, q) 平面中的分类。
| Condition | 条件 | Eigenvalues | 特征值 | Type | 类型 | Stability | 稳定性 |
| q < 0 | Real, opposite signs (λ₁ < 0 < λ₂) | Saddle point | Unstable | 不稳定 |
| q > 0, p² > 4q, p > 0 | Real, both positive | Unstable node | Unstable | 不稳定 |
| q > 0, p² > 4q, p < 0 | Real, both negative | Stable node | Asymptotically stable | 渐近稳定 |
| q > 0, p² < 4q, p > 0 | Complex, Re(λ) > 0 | Unstable spiral | Unstable | 不稳定 |
| q > 0, p² < 4q, p < 0 | Complex, Re(λ) < 0 | Stable spiral | Asymptotically stable | 渐近稳定 |
| q > 0, p = 0 | Purely imaginary | Centre | Stable | 稳定 |
| q = p²/4, p > 0 | Repeated positive | Degenerate node | Unstable | 不稳定 |
| q = p²/4, p < 0 | Repeated negative | Degenerate node | Asymptotically stable | 渐近稳定 |
Notice that when q > 0 and p > 0, regardless of whether eigenvalues are real or complex, the equilibrium point is unstable. When q > 0 and p < 0, it is asymptotically stable. Therefore, the sign of p tells us about stability when det A > 0.
注意到当 q > 0 且 p > 0 时，无论特征值是实数还是复数，平衡点都不稳定。当 q > 0 且 p < 0 时，则渐近稳定。因此，当 det A > 0 时，p 的符号告诉我们稳定性。
5. Stable Node | 稳定结点
When both eigenvalues are real, distinct, and negative (λ₁ < λ₂ < 0), the origin is a stable node. All solution trajectories approach the origin as t increases, tangent to the eigenvector corresponding to the eigenvalue with the smaller absolute value (the slower decay).
当两个特征值都是实数、互异且为负（λ₁ < λ₂ < 0）时，原点是一个稳定结点。所有解轨线都随着 t 的增大而趋近原点，并与对应绝对值较小特征值（衰减较慢的那个）的特征向量相切。
An example is the system dX/dt = A·X with A = [−3 0; 0 −1]. The eigenvalues are −3 and −1. Solutions are of the form x = C₁e⁻³ᵗ, y = C₂e⁻ᵗ. Since e⁻³ᵗ decays faster than e⁻ᵗ, the trajectory approaches the origin along the y-axis direction near the origin.
一个例子是系统 dX/dt = A·X，其中 A = [−3 0; 0 −1]。特征值为 −3 和 −1。解的形式为 x = C₁e⁻³ᵗ，y = C₂e⁻ᵗ。由于 e⁻³ᵗ 比 e⁻ᵗ 衰减得更快，轨线在原点附近沿 y 轴方向趋近原点。
Visually, the phase portrait shows all arrows pointing inward toward the origin. The trajectories look like parabolas bending toward the eigenvector with the slower decay rate. For IB exams, you should be able to sketch this phase portrait given the eigenvalues and eigenvectors.
从视觉上看，相图中所有箭头都指向原点。轨线看起来像抛物线，向衰减较慢的特征向量方向弯曲。对于IB考试，你应该能够根据给定的特征值和特征向量画出这个相图。
6. Unstable Node | 不稳定结点
When both eigenvalues are real, distinct, and positive (0 < λ₁ < λ₂), the origin is an unstable node. This is the time-reversed version of the stable node: trajectories move away from the origin in all directions.
当两个特征值都是实数、互异且为正（0 < λ₁ < λ₂）时，原点是不稳定结点。这是稳定结点的时间反转版本：轨线在所有方向上远离原点。
Consider A = [2 0; 0 5]. The eigenvalues are 2 and 5, with corresponding eigenvectors along the x and y axes. Solutions blow up as t increases: x = C₁e²ᵗ, y = C₂e⁵ᵗ. The trajectory is pulled away more rapidly along the direction of the larger eigenvalue.
考虑 A = [2 0; 0 5]。特征值为 2 和 5，对应的特征向量沿 x 轴和 y 轴方向。随着 t 增大解会爆炸式增长：x = C₁e²ᵗ，y = C₂e⁵ᵗ。轨线沿较大特征值方向被更快地拉离。
In the phase portrait, all trajectories originate from the origin (as t → −∞) and move outward, eventually aligning with the eigenvector corresponding to the larger eigenvalue. The origin is repelling in every direction.
在相图中，所有轨线都源自原点（当 t → −∞ 时）并向外移动，最终与较大特征值对应的特征向量对齐。原点在所有方向上都是排斥的。
7. Saddle Point | 鞍点
A saddle point occurs when the eigenvalues are real with opposite signs, i.e., λ₁ < 0 < λ₂. The name comes from the shape of the phase portrait, resembling the contours of a horse's saddle. The origin is neither a source nor a sink but a combination of both.
当特征值为实数且符号相反时，即 λ₁ < 0 < λ₂，就出现鞍点。这个名称来源于相图的形状，类似于马鞍的轮廓。原点既不是源也不是汇，而是两者的组合。
Let us examine the system with A = [1 0; 0 −2]. Eigenvalues are 1 and −2. The x-direction is repelling (unstable manifold: the line spanned by the eigenvector for λ = 1), while the y-direction is attracting (stable manifold: the line spanned by the eigenvector for λ = −2).
让我们考察 A = [1 0; 0 −2] 的系统。特征值为 1 和 −2。x 方向是排斥的（不稳定流形：由 λ = 1 的特征向量张成的直线），而 y 方向是吸引的（稳定流形：由 λ = −2 的特征向量张成的直线）。
Most trajectories approach the origin briefly along the stable manifold and then veer off along the unstable manifold toward infinity. Only solutions starting exactly on the stable manifold converge to the origin as t → ∞.
大多数轨线先沿稳定流形短暂地接近原点，然后沿不稳定流形转向无穷远。只有恰好从稳定流形上出发的解才能在 t → ∞ 时收敛到原点。
For IB, the key exam point is to remember: a saddle point is always unstable, even though half of the phase space moves toward the origin. The existence of any direction along which solutions escape implies instability.
对于IB考试，关键考点是记住：鞍点总是不稳定的，即使一半的相空间向原点移动。只要存在任何一个方向让解逃逸，就意味着不稳定。
8. Spiral Points | 螺旋点
When the eigenvalues are complex conjugates λ = p ± iω with ω ≠ 0, the solutions involve oscillations. The real part p determines whether the oscillations grow, decay, or maintain constant amplitude. This gives rise to spiral trajectories in the phase plane.
当特征值为共轭复数 λ = p ± iω（ω ≠ 0）时，解涉及振荡。实部 p 决定了振荡是增长、衰减还是保持恒定振幅。这就在相平面中产生了螺旋形轨线。
For p < 0, we have a stable spiral: trajectories spiral into the origin as t → ∞. For p > 0, we have an unstable spiral: trajectories spiral outward away from the origin. For p = 0, we obtain a centre: trajectories form closed elliptical orbits around the origin.
当 p < 0 时，为稳定螺旋：轨线随着 t → ∞ 螺旋进入原点。当 p > 0 时，为不稳定螺旋：轨线螺旋向外远离原点。当 p = 0 时，得到中心：轨线围绕原点形成闭合的椭圆轨道。
An example of a stable spiral is given by A = [−1 2; −2 −1]. The eigenvalues are −1 ± 2i. The angular frequency ω = 2 determines how quickly the trajectory winds around the origin, while p = −1 determines the exponential decay of the amplitude.
稳定螺旋的一个例子是 A = [−1 2; −2 −1]。特征值为 −1 ± 2i。角频率 ω = 2 决定了轨线围绕原点缠绕的速度，而 p = −1 决定了振幅的指数衰减。
The direction of rotation (clockwise or counterclockwise) can be determined by testing a single point. For instance, at the point (1, 0), compute dy/dt = cx + dy. If dy/dt < 0, the trajectory moves downward initially, indicating clockwise rotation.
旋转方向（顺时针或逆时针）可以通过测试一个点来确定。例如，在点 (1, 0) 处，计算 dy/dt = cx + dy。如果 dy/dt < 0，则轨线最初向下移动，表明是顺时针旋转。
9. Centre | 中心
A centre occurs when the eigenvalues are purely imaginary, λ = ±iω. In this case, the solutions are periodic functions, and the trajectories are closed curves (ellipses or circles) surrounding the origin. The origin is stable but not asymptotically stable.
当特征值为纯虚数 λ = ±iω 时，出现中心。在这种情况下，解是周期函数，轨线是围绕原点的闭合曲线（椭圆或圆）。原点是稳定的，但不是渐近稳定的。
The canonical example is the simple harmonic oscillator: dx/dt = y, dy/dt = −x. The matrix A = [0 1; −1 0] has eigenvalues ±i. The solutions are x = C₁cos(t) + C₂sin(t), y = −C₁sin(t) + C₂cos(t), which trace out circles.
典型例子是简谐振子：dx/dt = y，dy/dt = −x。矩阵 A = [0 1; −1 0] 的特征值为 ±i。解为 x = C₁cos(t) + C₂sin(t)，y = −C₁sin(t) + C₂cos(t)，它们描绘出圆周。
A centre is extremely sensitive: any small perturbation to the matrix that gives the eigenvalues a non-zero real part will turn the closed orbits into spirals, changing the stability qualitatively. Thus, centres are a critical boundary case in the classification scheme.
中心是极其敏感的：对矩阵的任何微小扰动，若使特征值具有非零实部，都会将闭合轨道变成螺旋，从而在本质上改变稳定性。因此，中心是分类方案中的一个关键的边界情况。
10. Degenerate Nodes | 退化结点
When the discriminant Δ = 0, there is a single repeated eigenvalue λ. The behaviour depends on whether the matrix A is diagonalisable or not. This gives rise to degenerate or star nodes.
当判别式 Δ = 0 时，存在一个重特征值 λ。系统的行为取决于矩阵 A 是否可对角化。这就产生了退化结点或星形结点。
If A is diagonalisable (i.e., there are two linearly independent eigenvectors), the phase portrait consists of straight lines radiating from the origin, all with the same slope pattern. This is called a star node. If A is defective (only one eigenvector), trajectories approach the origin tangent to the single eigenvector, forming a degenerate node.
如果 A 可对角化（即有两个线性无关的特征向量），相图由从原点辐射出的直线组成，具有相同的斜率模式。这称为星形结点。如果 A 有缺陷（只有一个特征向量），轨线沿唯一特征向量的方向趋近原点，形成退化结点。
For example, A = [−1 0; 0 −1] has a repeated eigenvalue −1 and is diagonalisable, producing a star node. Meanwhile, A = [−1 1; 0 −1] has eigenvalue −1 with algebraic multiplicity 2 but geometric multiplicity 1, producing a degenerate node with trajectories curving into the origin along the x-axis.
例如，A = [−1 0; 0 −1] 具有重特征值 −1 且可对角化，产生星形结点。而 A = [−1 1; 0 −1] 的特征值 −1 的代数重数为 2 但几何重数为 1，产生退化结点，轨线沿 x 轴弯曲进入原点。
In IB exams, it is crucial to check the geometric multiplicity when you encounter a repeated eigenvalue. Remember: repeated eigenvalue with n independent eigenvectors → star node; with fewer than n eigenvectors → degenerate node.
在IB考试中，遇到重特征值时务必检查几何重数。记住：重特征值对应 n 个独立特征向量 → 星形结点；少于 n 个特征向量 → 退化结点。
11. Worked Example | 例题详解
Let us apply the classification method to a concrete example. Classify the equilibrium at the origin for the system:
让我们将分类方法应用于一个具体例子。对以下系统，分类原点的平衡点：
dx/dt = 3x − y, dy/dt = 2x + y
Here A = [3 −1; 2 1]. First, compute the trace p = 3 + 1 = 4 and the determinant q = (3)(1) − (−1)(2) = 5. Since q = 5 > 0 and p = 4 > 0, we know the equilibrium is unstable. To determine the specific type, compute the discriminant Δ = p² − 4q = 16 − 20 = −4 < 0.
此处 A = [3 −1; 2 1]。首先计算迹 p = 3 + 1 = 4，行列式 q = (3)(1) − (−1)(2) = 5。由于 q = 5 > 0 且 p = 4 > 0，我们知道平衡点是不稳定的。为确定具体类型，计算判别式 Δ = p² − 4q = 16 − 20 = −4 < 0。
Since Δ < 0 and p > 0 with q > 0, the origin is an unstable spiral. The eigenvalues are λ = (4 ± 2i)/2 = 2 ± i. The positive real part confirms the spiralling-out behaviour. If asked to sketch the phase portrait, draw an anticlockwise spiral moving outward (verify by checking the sign of dy/dt at (1, 0)).
由于 Δ < 0 且 p > 0，q > 0，原点是不稳定螺旋。特征值为 λ = (4 ± 2i)/2 = 2 ± i。正的实部确认了螺旋发散的行为。如果被要求画出相图草图，画出逆时针向外螺旋的轨迹（通过在点 (1, 0) 处检查 dy/dt 的正负来验证方向）。
This demonstrates the power of the classification scheme: with just two quick computations (trace and determinant), we can determine the qualitative behaviour without ever solving the system explicitly.
这展示了分类方案的力量：仅需两个快速计算（迹和行列式），我们就能确定定性的行为，而无需显式求解系统。
12. Summary and Exam Tips | 总结与考试技巧
The classification of linear equilibrium points is a systematic process. First, compute the eigenvalues of the matrix A. Then, apply the following decision tree: (1) Real eigenvalues of opposite signs → saddle (unstable); (2) Real eigenvalues both negative → stable node; (3) Real eigenvalues both positive → unstable node; (4) Complex eigenvalues with negative real part → stable spiral; (5) Complex eigenvalues with positive real part → unstable spiral; (6) Purely imaginary eigenvalues → centre (stable).
线性平衡点的分类是一个系统化的过程。首先，计算矩阵 A 的特征值。然后应用如下决策树：(1) 异号实特征值 → 鞍点（不稳定）；(2) 实特征值均为负 → 稳定结点；(3) 实特征值均为正 → 不稳定结点；(4) 实部为负的复特征值 → 稳定螺旋；(5) 实部为正的复特征值 → 不稳定螺旋；(6) 纯虚特征值 → 中心（稳定）。
For the (p, q)-plane approach: if q < 0, it is always a saddle. If q > 0, the sign of p determines stability: p > 0 means unstable, p < 0 means asymptotically stable. Then use the discriminant to distinguish nodes from spirals. This shortcut can save valuable time in exams.
对于 (p, q) 平面的方法：若 q < 0，则一定是鞍点。若 q > 0，p 的符号决定稳定性：p > 0 意味着不稳定，p < 0 意味着渐近稳定。然后用判别式区分结点和螺旋。这种捷径可以在考试中节省宝贵时间。
Common IB pitfalls to avoid: confusing “stable” with “asymptotically stable” (a centre is stable but not asymptotically); forgetting that a saddle is unstable even though some trajectories approach; losing marks by not indicating the direction of flow in phase portraits; and incorrectly drawing the tangent lines at the origin for nodes and saddles.
需要避免的常见IB陷阱：混淆”稳定”与”渐近稳定”（中心是稳定但非渐近稳定的）；忘记鞍点即使有部分轨线趋近但整体不稳定的性质；在相图中不标注流的方向而失分；以及在原点处错误地画出结点和鞍点的切线方向。
Finally, remember that these classifications apply to linear systems. For non-linear systems, the Hartman-Grobman theorem assures us that the local behaviour near a hyperbolic equilibrium point (where eigenvalues have non-zero real parts) is qualitatively the same as that of its linearisation. This justifies using linear classification for non-linear systems — a topic often examined in IB extension questions.
最后，请记住这些分类适用于线性系统。对于非线性系统，Hartman-Grobman 定理保证了在双曲平衡点（特征值实部不为零）附近的局部行为与其线性化系统的行为在定性上相同。这为使用线性分类处理非线性系统提供了理论依据——这也是IB扩展题目中常考的内容。
Published by TutorHao | Mathematics Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
屏轩国际教育cambridge primary/secondary checkpoint, cat4, ukiset,ukcat,igcse,alevel,PAT,STEP,MAT, ibdp,ap,ssat,sat,sat2课程辅导，国外大学本科硕士研究生博士课程论文辅导