A-Level CIE Further Mathematics: Practical Guide to Numerical Experiments | CIE A-Level 进阶数学：数值实验操作指南

📚 A-Level CIE Further Mathematics: Practical Guide to Numerical Experiments | CIE A-Level 进阶数学：数值实验操作指南

Numerical methods form a vital part of CIE A-Level Further Mathematics, empowering students to solve equations, evaluate integrals, and approximate differential equations through step-by-step computational algorithms. This guide walks you through the core techniques as if performing controlled numerical experiments — emphasising implementation, error awareness, and practical tips that matter in the exam. Each section pairs the theoretical underpinning with a hands-on recipe, so you can carry out your own ‘experiments’ with confidence.

数值方法是 CIE A-Level 进阶数学的重要组成部分，让学生能够通过逐步的计算算法求解方程、估算积分和近似微分方程。本指南如同进行受控的数值实验一样，带你掌握核心技术，强调实施方法、误差意识以及考试中至关重要的实用技巧。每一节都将理论基础与动手步骤相结合，让你有能力自信地完成自己的“实验”。

1. Introduction to Numerical Experiments | 数值实验导论

In Further Mathematics, a ‘numerical experiment’ refers to the systematic application of an algorithm to obtain an approximate solution where analytic methods fail. You set up an initial value, iterate using a prescribed formula, observe convergence, and estimate error — just as in a laboratory investigation.

在进阶数学中，“数值实验”指的是系统地应用算法来获取解析方法无法求解的近似解。你设定初始值，使用给定公式迭代，观察收敛情况并估计误差——就像在实验室进行探究一样。

The CIE syllabus explicitly expects you to derive iterative formulas, use them with a given number of iterations, and discuss the reliability of the result. Treat each question as a mini-experiment: document your steps, manage precision, and always check whether the method is appropriate for the function’s behaviour.

CIE 大纲明确要求你会推导迭代公式，在给定迭代次数下使用它们，并讨论结果的可靠性。将每一道题视为一个小型实验：记录步骤，管理精度，并始终检查方法是否适合函数的行为。

2. Understanding Iterative Formulas and Convergence Criteria | 理解迭代公式与收敛准则

An iterative formula produces a sequence x₀, x₁, x₂, … that hopefully converges to a root α. The general form is xₙ₊₁ = g(xₙ). Convergence is guaranteed near α if |g'(α)| < 1. This is the core condition you should test before starting the iteration.

迭代公式产生一个序列 x₀, x₁, x₂, …，希望它收敛到根 α。一般形式为 xₙ₊₁ = g(xₙ)。如果 |g'(α)| < 1，则在 α 附近收敛得到保证。这是你在开始迭代前应检验的核心条件。

When you choose an initial approximation, sketch the function to see the behaviour. Monotonic convergence, oscillatory convergence, and divergence can be predicted from the sign and magnitude of g'(x). Always ask: is my rearrangement likely to converge?

当你选择初始近似值时，画出函数草图以观察其行为。单调收敛、振荡收敛和发散可以通过 g'(x) 的符号和大小进行预测。始终要问：我构造的迭代式是否可能收敛？

3. The Newton-Raphson Method: Quadratic Convergence | 牛顿-拉弗森方法：二次收敛

The Newton-Raphson formula is xₙ₊₁ = xₙ − f(xₙ)/f'(xₙ). It is one of the most efficient root-finding ‘experiments’ because it exhibits quadratic convergence near a simple root, meaning the number of correct digits roughly doubles each step.

牛顿-拉弗森公式为 xₙ₊₁ = xₙ − f(xₙ)/f'(xₙ)。它是最高效的寻根“实验”之一，因为在单根附近表现出二次收敛，即正确数字的位数每一步大约翻倍。

To apply it, you must be able to differentiate f(x) and evaluate both f and f’ at each step. Begin with a value x₀ close to the root, ideally where f(x₀) and f'(x₀) have the same sign to avoid oscillation. In the exam, you are often given the derivative, but you must confirm it is correct.

要应用该方法，你必须能对 f(x) 求导并在每一步计算 f 和 f’ 的值。从一个靠近根的 x₀ 开始，理想情况下 f(x₀) 和 f'(x₀) 同号，以避免振荡。考试中通常会给出导数，但你仍需确认其正确性。

Beware of stationary points or near-zero derivatives, which can send the next iterate far away. In such cases, switch to a more robust method or refine your starting value. Always record f(xₙ) to track the decrease in residual.

小心驻点或接近零的导数，这会使下一个迭代值跳得很远。遇到这种情况，应切换到更稳健的方法或改进初始值。始终记录 f(xₙ) 以追踪残差的减小。

4. Fixed-Point Iteration: Rearranging f(x)=0 into x=g(x) | 不动点迭代：将 f(x)=0 改写为 x=g(x)

A simpler experimental approach is fixed-point iteration: rearrange f(x)=0 to the form x = g(x) and use xₙ₊₁ = g(xₙ). The root α satisfies α = g(α), hence the term ‘fixed point’. The choice of rearrangement is critical for success.

一种更简单的实验方法是不动点迭代：将 f(x)=0 改写为 x = g(x) 的形式，并使用 xₙ₊₁ = g(xₙ)。根 α 满足 α = g(α)，因此叫做“不动点”。改写方式的选择对于成功至关重要。

For example, x³ − 2x − 5 = 0 can be rearranged to x = ∛(2x + 5) or x = (x³ − 5)/2. The first rearrangement yields |g'(α)| < 1, while the second diverges. Always calculate g'(x) and evaluate near the estimated root to predict behaviour.

例如，x³ − 2x − 5 = 0 可改写为 x = ∛(2x + 5) 或 x = (x³ − 5)/2。第一种改写使 |g'(α)| < 1，而第二种发散。永远要计算 g'(x) 并在估计的根附近求值以预测其行为。

In the experiment, draw a cobweb diagram using the line y = x and the curve y = g(x). Starting from x₀, move vertically to the curve, horizontally to the line, and repeat. This visual tool helps you see convergence, staircase patterns, and spiralling.

在实验中，使用直线 y = x 和曲线 y = g(x) 绘制楼梯图。从 x₀ 开始，垂直移动到曲线，水平移动到直线，如此重复。这个可视化工具可帮助你观察收敛、阶梯模式和螺旋现象。

5. The Trapezium Rule: Estimating Definite Integrals | 梯形法则：估算定积分

When you cannot integrate analytically, the trapezium rule offers a straightforward numerical integration experiment. Divide the interval [a, b] into n strips of equal width h = (b-a)/n, and compute ∫ₐᵇ f(x) dx ≈ h/2 [y₀ + 2(y₁ + y₂ + … + yₙ₋₁) + yₙ], where yᵢ = f(xᵢ).

当你无法解析积分时，梯形法则提供了一个直接了当的数值积分实验。将区间 [a, b] 等分为 n 个宽度为 h = (b-a)/n 的窄条，计算 ∫ₐᵇ f(x) dx ≈ h/2 [y₀ + 2(y₁ + y₂ + … + yₙ₋₁) + yₙ]，其中 yᵢ = f(xᵢ)。

Increasing n generally improves accuracy, but the error is proportional to h² and depends on f”(x). The method works by approximating the curve with straight line segments, so it is exact only for linear functions. When the curve is highly concave or convex, the error can be noticeable even with many strips.

增加 n 通常会提高精确度，但误差与 h² 成正比，并依赖于 f”(x)。该方法用直线段近似曲线，因此只对线性函数是精确的。当曲线凹或凸的程度很高时，即使使用很多窄条，误差也可能显著。

For a numerical experiment, you might be asked to use a given number of ordinates or to double the number of strips and compare results. Always tabulate x and y values carefully, and use a systematic recording method to avoid arithmetic slips.

在数值实验中，你可能被要求使用给定数量的纵坐标，或将窄条数量加倍并比较结果。务必仔细列表记录 x 和 y 的值，并使用系统化的记录方法以避免算术错误。

6. Simpson’s Rule: Parabolic Approximation for Better Accuracy | 辛普森法则：抛物线近似提高精度

Simpson’s rule improves upon the trapezium rule by fitting quadratic arcs through sets of three consecutive points. For n even strips, Simpson’s rule gives ∫ₐᵇ f(x) dx ≈ h/3 [y₀ + 4(y₁ + y₃ + … + yₙ₋₁) + 2(y₂ + y₄ + … + yₙ₋₂) + yₙ]. The coefficients follow a 1-4-2-4-2…-4-1 pattern.

辛普森法则通过用二次弧线拟合每三个连续点，改进了梯形法则。对于偶数 n 条窄条，辛普森法则给出 ∫ₐᵇ f(x) dx ≈ h/3 [y₀ + 4(y₁ + y₃ + … + yₙ₋₁) + 2(y₂ + y₄ + … + yₙ₋₂) + yₙ]。系数遵循 1-4-2-4-2…-4-1 的规律。

This method is often much more accurate for smooth functions because the error term is proportional to h⁴, provided the fourth derivative is continuous. When you perform this experiment, notice how doubling n produces a far greater improvement than with the trapezium rule.

对于光滑函数，该方法的精确度往往高得多，因为误差项与 h⁴ 成正比，前提是四阶导数连续。当你进行这个实验时，会注意到加倍 n 带来的改进远大于梯形法则。

Because the weights differ, never confuse the 1-4-2 pattern with the trapezium rule’s 1-2-2…-2-1 sequence. A practical tip: highlight the odd and even ordinates in different colours in your table to avoid weighting errors during the calculation.

由于权重不同，切勿将 1-4-2 的规律与梯形法则的 1-2-2…-2-1 序列混淆。实用提示：在表格中用不同颜色标示奇数和偶数纵坐标，以避免计算中的权重错误。

7. Euler’s Method: First-Order Numerical Solution of ODEs | 欧拉方法：常微分方程的一阶数值解

Euler’s method provides a simple recipe for the numerical integration of a first-order differential equation dy/dx = f(x, y) with initial condition y(x₀) = y₀. Step forward using yₙ₊₁ = yₙ + h f(xₙ, yₙ), where h is the step size, and xₙ₊₁ = xₙ + h.

欧拉方法为数值积分一阶微分方程 dy/dx = f(x, y) 及初始条件 y(x₀) = y₀ 提供了一个简单的方法。利用 yₙ₊₁ = yₙ + h f(xₙ, yₙ) 向前推进，其中 h 为步长，且 xₙ₊₁ = xₙ + h。

Think of the method as constructing a polygonal path that follows the local tangent. The accuracy is only first order, meaning the global error is proportional to h. Reducing h by a factor of 10 reduces the error roughly by a factor of 10, but increases the number of steps and rounding errors.

可以将这个方法想象成构建一条沿着局部切线的折线路径。其精度仅为一阶，意味着全局误差与 h 成正比。将 h 减小为原来的十分之一会大致使误差减小为十分之一，但会增加步数和舍入误差。

In your experiment, set up a table with columns for n, xₙ, yₙ, f(xₙ, yₙ), and the next y. Work systematically and retain more decimal places during intermediate steps than the final answer requires — this is a standard experimental discipline.

在实验中，建立一个包含 n、xₙ、yₙ、f(xₙ, yₙ) 和下一个 y 值列的表。系统地计算，并在中间步骤中保留比最终答案所需更多的小数位——这是一种标准的实验规范。

8. Improved Euler Method (Heun’s Method): Reducing Error | 改进欧拉法（休恩法）：减小误差

To increase the order of accuracy while keeping the experiment manageable, the Improved Euler method uses a predictor-corrector framework. First, a preliminary step (predictor) y* = yₙ + h f(xₙ, yₙ) is taken, then corrected using the average slope: yₙ₊₁ = yₙ + h/2 [f(xₙ, yₙ) + f(xₙ₊₁, y*)].

为在保持实验可操作的同时提高精度阶数，改进欧拉法采用了预估-校正框架。首先，执行一个初步步（预估步）y* = yₙ + h f(xₙ, yₙ)，然后使用平均斜率进行校正：yₙ₊₁ = yₙ + h/2 [f(xₙ, yₙ) + f(xₙ₊₁, y*)]。

This method has second-order accuracy, which means halving h reduces the error by roughly a factor of 4. The extra work pays off when you need a more reliable numerical trajectory. In exam contexts, you might be asked to apply both Euler and Improved Euler to a given ODE and compare the results.

该方法具有二阶精度，这意味着步长减半可使误差大约减小为四分之一。当你需要更可靠的数值轨迹时，额外的工作会带来回报。在考试情境中，你可能被要求对给定常微分方程同时应用欧拉法和改进欧拉法并比较结果。

When conducting the experiment, always record the predictor value separately to avoid confusing it with the corrected value. A well-organised table with sub-columns for the predictor and the corrected slope makes the process transparent and reduces mistakes.

进行实验时，始终单独记录预估步值，以免与校正值混淆。一个包含预估步和校正斜率子列的结构清晰的表格可使过程透明并减少错误。

9. Error Bounds and Practical Error Estimation | 误差界与实用误差估计

Every numerical experiment must include an error discussion. For the trapezium rule, the error E satisfies |E| ≤ K(b-a)h²/12 where K is an upper bound for |f”(x)| on [a,b]. For Simpson’s rule, |E| ≤ M(b-a)h⁴/180 with M bounding the fourth derivative.

每个数值实验都必须包含误差讨论。对于梯形法则，误差 E 满足 |E| ≤ K(b-a)h²/12，其中 K 是 |f”(x)| 在 [a,b] 上的一个上界。对于辛普森法则，|E| ≤ M(b-a)h⁴/180，M 为四阶导数的界。

In exam problems, you often calculate an integral with different step sizes and use the difference between two approximations to estimate the true error. This hands-on approach mirrors real experimental data analysis: compare results from n strips and 2n strips and infer the convergence rate.

在考试题中，你常会用不同步长计算积分，并利用两个近似值之间的差异来估计真实误差。这种动手方法反映了真实的实验数据分析：比较 n 条和 2n 条窄条的结果，并推断收敛速率。

For iterative root-finding, the change between successive iterates Δx = |xₙ₊₁ – xₙ| is a simple error indicator. As convergence sets in, Δx becomes smaller than the true error. When recording your experimental results, always note the stopping criterion — e.g., the process stopped when |xₙ₊₁ – xₙ| < 0.0005.

对于迭代寻根，相邻两次迭代值的变化 Δx = |xₙ₊₁ – xₙ| 是一个简单的误差指标。随着收敛的发生，Δx 会变得小于真实误差。记录实验结果时，务必注明停止准则——例如，当 |xₙ₊₁ – xₙ| < 0.0005 时停止迭代。

10. Numerical Stability and Choosing Step Size | 数值稳定性与步长选择

In differential equation experiments, stability is just as important as accuracy. Euler’s method can become unstable if h is too large relative to the stiffness of the equation. For dy/dx = −λy (λ > 0), the numerical solution blows up unless h < 2/λ, even though the true solution decays exponentially.

在微分方程实验中，稳定性与精确度同等重要。如果步长 h 相对于方程的刚性过大，欧拉方法可能变得不稳定。对于 dy/dx = −λy (λ > 0)，除非 h < 2/λ，否则数值解会暴涨，而真实解却呈指数衰减。

Choosing the right h is an experimental design decision. Start with a moderate step size, observe whether the solution behaves physically (e.g., stays positive, remains bounded), and then halve h to check if the results converge to a consistent trajectory. This is the essence of a numerical convergence study.

选择合适的 h 是一项实验设计决策。从中等步长开始，观察解的表现是否符合物理规律（如保持正值、有界），然后将步长减半以检验结果是否收敛到一致的轨迹。这就是数值收敛研究的本质。

For integral approximations, the choice of n often depends on the required precision. Sometimes the question dictates the number of strips; at other times you must justify why a particular n gives the desired accuracy using the error bound formulas. Always show the reasoning.

对于积分近似，n 的选择常取决于所需精度。有时题目会指定窄条数量；有时你必须利用误差界公式证明为何特定的 n 能达到所需精度。始终要展示推理过程。

11. Using Technology: Spreadsheets and Programmable Calculators | 运用技术：电子表格与可编程计算器

Modern numerical experiments in A-Level Further Mathematics are greatly accelerated by technology, but you must still understand the underlying algorithm. A spreadsheet can be set up to perform iterations and graph convergence, while a programmable calculator allows you to run Euler’s method efficiently in the exam.

A-Level 进阶数学中的现代数值实验可通过技术大大加速，但你仍须理解底层算法。电子表格可设置为执行迭代并绘制收敛图，而可编程计算器则允许你在考试中高效运行欧拉方法。

However, never rely blindly on the output. Cross-check the first couple of iterations by hand to ensure the formula is entered correctly. The exam may ask you to demonstrate the calculation step by step; the calculator is a verification tool, not a replacement for understanding.

然而，切勿盲目信赖输出结果。手工核对最初几次迭代，以确保公式输入正确。考试可能要求你逐步展示计算过程；计算器只是验证工具，不能替代理解。

When using a spreadsheet for a trapezium rule experiment, use absolute cell references for h and y-values. This mimics good lab practice where parameters are stored in designated cells, making it easy to adjust the number of strips and observe the effect on the integral.

当使用电子表格进行梯形法则实验时，对 h 和 y 值运用绝对单元格引用。这模仿了良好的实验室实践，即参数存储在指定单元格中，便于调整窄条数量并观察对积分的影响。

12. Exam Strategy and Common Mistakes | 考试策略与常见错误

In the CIE Further Mathematics exam, numerical methods questions are highly structured and reward careful, systematic work. Write the iterative formula clearly at the start of each part, and present your results in a neat table. This not only organises your thinking but also earns method marks, even if a later arithmetic error creeps in.

在 CIE 进阶数学考试中，数值方法题目结构性强，奖励谨慎、系统的工作。在每个部分开头清晰写出迭代公式，并用整洁的表格呈现结果。这不仅组织思路，还能在后续出现算术错误时赢得方法分。

Common pitfalls include: forgetting to check the convergence criterion before starting an iteration; mixing up ordinate weights in Simpson’s rule; using radians when degrees are intended in trigonometric integrations (or vice versa); and rounding intermediate values too early. Treat every calculation as part of a replicable experiment — record raw values, then round only at the final answer.

常见误区包括：开始迭代前忘记检查收敛准则；混淆辛普森法则中的纵坐标权重；在三角函数积分时将弧度与度数用混；以及过早地对中间值四舍五入。将每次计算视为可重复的实验——记录原始值，仅在最终答案处进行舍入。

Also, be prepared to compare methods: the question might ask whether Newton-Raphson or fixed-point iteration is more suitable for a given function. Justify your choice using |g'(α)| or by considering the ease of differentiation. When discussing error, always refer to the appropriate error term and support your arguments with calculations.

此外，应准备好比较不同方法：题目可能询问对于给定函数，牛顿法与不动点迭代哪个更适用。使用 |g'(α)| 或考虑求导的难易程度来论证你的选择。在讨论误差时，始终引用适当的误差项并用计算支撑你的论点。

Finally, view every numerical experiment as an opportunity to demonstrate not just your computational skills, but also your understanding of stability, convergence, and the limitations of the algorithms. This mature perspective is exactly what examiners look for in high-scoring scripts.

最后，把每个数值实验都视为一个机会，不仅展示你的计算技能，还展现你对算法稳定性、收敛性及其局限性的理解。这种成熟的视角正是考官在高分答卷中所寻找的。

Published by TutorHao | Further Mathematics Revision Series | aleveler.com

更多咨询请联系16621398022（同微信）