← All Units← 返回单元列表 ← Course Hub← 课程主页
U N I V E R S I T Y  C A L C U L U S
Unit A6 · Solutions解答

Linear Approximation and L'Hopital's Rule · Solutions线性近似与洛必达法则 · 解答

Companion to the University-Style Practice Set大学风格练习题配套解答

MEDIUM HARD CORE PROOF APPLIED

Sections 1 to 7: linearization, differentials and error estimation, Newton's method, indeterminate forms, L'Hopital's rule, exponential indeterminate forms, and growth rates1 至 7 节:线性化、微分与误差估计、牛顿迭代法、不定式、洛必达法则、指数不定式及增长率CALC I



PART I  ·  CORE TECHNIQUES第一部分  ·  核心技法Computational fluency · 28 marks计算熟练度 · 28 分

Worked Solutions详细解答

Q1MEDIUMCORElinearization and numerical approximation线性化与数值近似[6 marks]

Approximate (a) $\sqrt{4.1}$; (b) $\sqrt[3]{8.1}$; (c) $\sin(0.1)$ using linearization at a convenient base point.用线性化在适当基点处近似计算 (a) $\sqrt{4.1}$;(b) $\sqrt[3]{8.1}$;(c) $\sin(0.1)$。

Answers:答案:  (a) $2.025$  ·  (b) $2+\tfrac{1}{120}\approx 2.0083$  ·  (c) $0.1$

(a) sqrt(4.1): linearize f(x) = sqrt(x) at a = 4(a) $\sqrt{4.1}$:在 $a=4$ 处对 $f(x)=\sqrt{x}$ 线性化 M1·A1

Choose $f(x)=\sqrt{x}$, $a=4$. Then $f'(x)=\dfrac{1}{2\sqrt{x}}$, so $f'(4)=\dfrac{1}{4}$. (M1)取 $f(x)=\sqrt{x}$,$a=4$。则 $f'(x)=\dfrac{1}{2\sqrt{x}}$,故 $f'(4)=\dfrac{1}{4}$。(M1)

The linearization is $L(x)=\sqrt{4}+\dfrac{1}{4}(x-4)=2+\dfrac{x-4}{4}$.线性化为 $L(x)=\sqrt{4}+\dfrac{1}{4}(x-4)=2+\dfrac{x-4}{4}$。

$$ L(4.1)=2+\frac{0.1}{4}=2+0.025=2.025. $$

The true value is $\sqrt{4.1}\approx 2.02485$, confirming the approximation is excellent. (A1)真实值为 $\sqrt{4.1}\approx 2.02485$,验证近似效果极佳。(A1)

(b) cbrt(8.1): linearize f(x) = x^{1/3} at a = 8(b) $\sqrt[3]{8.1}$:在 $a=8$ 处对 $f(x)=x^{1/3}$ 线性化 M1·A1

Choose $f(x)=x^{1/3}$, $a=8$. Then $f'(x)=\dfrac{1}{3}x^{-2/3}$, so $f'(8)=\dfrac{1}{3}\cdot\dfrac{1}{4}=\dfrac{1}{12}$. (M1)取 $f(x)=x^{1/3}$,$a=8$。则 $f'(x)=\dfrac{1}{3}x^{-2/3}$,故 $f'(8)=\dfrac{1}{3}\cdot\dfrac{1}{4}=\dfrac{1}{12}$。(M1)

The linearization is $L(x)=2+\dfrac{1}{12}(x-8)$.线性化为 $L(x)=2+\dfrac{1}{12}(x-8)$。

$$ L(8.1)=2+\frac{0.1}{12}=2+\frac{1}{120}\approx 2.0083. $$

(A1)

(c) sin(0.1): linearize f(x) = sin(x) at a = 0(c) $\sin(0.1)$:在 $a=0$ 处对 $f(x)=\sin x$ 线性化 M1·A1

Choose $f(x)=\sin x$, $a=0$. Then $f'(x)=\cos x$, so $f'(0)=1$. (M1)取 $f(x)=\sin x$,$a=0$。则 $f'(x)=\cos x$,故 $f'(0)=1$。(M1)

The linearization is $L(x)=\sin 0+1\cdot(x-0)=x$.线性化为 $L(x)=\sin 0+1\cdot(x-0)=x$。

$$ L(0.1)=0.1. $$

The true value is $\sin(0.1)\approx 0.09983$. (A1)真实值为 $\sin(0.1)\approx 0.09983$。(A1)

Insight.洞察。 The linearization $L(x)=f(a)+f'(a)(x-a)$ is the equation of the tangent line at $a$, used as a local stand-in for $f$. The closer $x$ is to $a$ and the smaller $|f''(a)|$ is, the tighter the approximation. The standard approximations $\sin\theta\approx\theta$, $e^{\theta}\approx 1+\theta$, and $(1+\theta)^{n}\approx 1+n\theta$ for small $\theta$ are all linearizations at $a=0$.线性化 $L(x)=f(a)+f'(a)(x-a)$ 就是 $f$ 在 $a$ 处的切线方程,作为 $f$ 的局部替代。$x$ 越接近 $a$、$|f''(a)|$ 越小,近似越精确。常用近似 $\sin\theta\approx\theta$、$e^{\theta}\approx 1+\theta$ 以及 $(1+\theta)^{n}\approx 1+n\theta$($\theta$ 较小时)均为 $a=0$ 处的线性化。
Q2MEDIUMCOREdifferentials and propagated error微分与误差传播[8 marks]

Sphere with $r=5$ cm, $|\Delta r|\le 0.04$ cm: (a) write $dV$; (b) estimate $|\Delta V|$; (c) find the maximum relative error as a percentage.球体 $r=5$ cm,$|\Delta r|\le 0.04$ cm:(a) 写出 $dV$;(b) 估计 $|\Delta V|$;(c) 求最大相对误差(百分比)。

Answers:答案:  (a) $dV=4\pi r^{2}\,dr$  ·  (b) $|\Delta V|\le 4\pi\approx 12.57\text{ cm}^{3}$  ·  (c) $2.4\%$

(a) The differential of volume(a) 体积的微分 M1·A1

Differentiate $V=\tfrac{4}{3}\pi r^{3}$ with respect to $r$: $\dfrac{dV}{dr}=4\pi r^{2}$. (M1) Hence $dV=4\pi r^{2}\,dr$. (A1)对 $V=\tfrac{4}{3}\pi r^{3}$ 关于 $r$ 求导:$\dfrac{dV}{dr}=4\pi r^{2}$。(M1) 故 $dV=4\pi r^{2}\,dr$。(A1)

(b) Maximum absolute error(b) 最大绝对误差 M1·M1·A1

The propagated error is approximated by $|\Delta V|\approx|dV|=4\pi r^{2}|dr|$. (M1)传播误差近似为 $|\Delta V|\approx|dV|=4\pi r^{2}|dr|$。(M1)

At $r=5$ and $|dr|=|\Delta r|\le 0.04$: (M1)当 $r=5$,$|dr|=|\Delta r|\le 0.04$ 时:(M1)

$$ |\Delta V|\le 4\pi(5)^{2}(0.04)=4\pi\cdot 25\cdot 0.04=4\pi\approx 12.57\text{ cm}^{3}. $$

(A1)

(c) Maximum relative error(c) 最大相对误差 M1·A1·A1

Divide $dV$ by $V$: (M1)用 $dV$ 除以 $V$:(M1)

$$ \frac{dV}{V}=\frac{4\pi r^{2}\,dr}{\tfrac{4}{3}\pi r^{3}}=\frac{3\,dr}{r}. $$

At $r=5$ and $|\Delta r|\le 0.04$: (A1)当 $r=5$,$|\Delta r|\le 0.04$ 时:(A1)

$$ \left|\frac{\Delta V}{V}\right|\le\frac{3\times 0.04}{5}=\frac{0.12}{5}=0.024=2.4\%. $$

(A1)

Insight.洞察。 The relative error formula $dV/V=3\,dr/r$ reveals a general pattern: for $V\propto r^{n}$, the relative error is $n$ times the relative error in the radius. Here $n=3$ (volume is cubic in $r$), so a $0.8\%$ error in $r$ $(0.04/5)$ produces a $2.4\%$ error in $V$. This multiplicative amplification is why precision in the measured variable is critical for higher-power quantities such as volumes, moments of inertia, and intensities.相对误差公式 $dV/V=3\,dr/r$ 揭示了一个普遍规律:对于 $V\propto r^{n}$,相对误差是半径相对误差的 $n$ 倍。这里 $n=3$(体积是 $r$ 的三次方),因此 $r$ 的 $0.8\%$ 误差($0.04/5$)导致 $V$ 产生 $2.4\%$ 的误差。这种倍增效应说明,对于体积、转动惯量、强度等高次幂量,测量变量的精度至关重要。
Q3HARDCOREL'Hopital's rule: 0/0 and infty/infty洛必达法则:$0/0$ 与 $\infty/\infty$ 型[8 marks]

Evaluate: (a) $\lim_{x\to 0}\frac{e^{x}-1-x}{x^{2}}$; (b) $\lim_{x\to\infty}\frac{x^{3}}{e^{x}}$; (c) $\lim_{x\to 0}\frac{\tan x-x}{x^{3}}$.求值:(a) $\lim_{x\to 0}\frac{e^{x}-1-x}{x^{2}}$;(b) $\lim_{x\to\infty}\frac{x^{3}}{e^{x}}$;(c) $\lim_{x\to 0}\frac{\tan x-x}{x^{3}}$。

Answers:答案:  (a) $\tfrac{1}{2}$  ·  (b) $0$  ·  (c) $\tfrac{1}{3}$

(a) Apply L'Hopital twice to a 0/0 form(a) 对 $0/0$ 型两次使用洛必达法则 M1·M1·A1

Direct substitution gives $\tfrac{0}{0}$. (M1) Apply L'Hopital once: numerator becomes $e^{x}-1$, denominator $2x$. At $x=0$ this is again $\tfrac{0}{0}$. (M1) Apply L'Hopital a second time:直接代入得 $\tfrac{0}{0}$。(M1) 使用一次洛必达法则:分子变为 $e^{x}-1$,分母为 $2x$。在 $x=0$ 处仍为 $\tfrac{0}{0}$。(M1) 再次使用洛必达法则:

$$ \lim_{x\to 0}\frac{e^{x}}{2}=\frac{e^{0}}{2}=\frac{1}{2}. $$

(A1)

(b) Apply L'Hopital three times to an infty/infty form(b) 对 $\infty/\infty$ 型三次使用洛必达法则 M1·M1·A1

The form is $\infty/\infty$. Differentiate numerator and denominator: (M1)极限形式为 $\infty/\infty$。对分子分母分别求导:(M1)

$$ \frac{x^{3}}{e^{x}}\xrightarrow{L'H}\frac{3x^{2}}{e^{x}}\xrightarrow{L'H}\frac{6x}{e^{x}}\xrightarrow{L'H}\frac{6}{e^{x}}\longrightarrow 0. $$

Each step maintains the $\infty/\infty$ form until the last, where the limit is $6/\infty=0$. (M1·A1)每步均保持 $\infty/\infty$ 形式,直至最后一步极限为 $6/\infty=0$。(M1·A1)

(c) Apply L'Hopital once, then use the standard limit(c) 使用一次洛必达法则,再利用标准极限 M1·A1

The form is $\tfrac{0}{0}$ at $x=0$. Differentiate: $\dfrac{\sec^{2}x-1}{3x^{2}}=\dfrac{\tan^{2}x}{3x^{2}}$. This is still $\tfrac{0}{0}$. (M1) Write在 $x=0$ 处形式为 $\tfrac{0}{0}$。求导得 $\dfrac{\sec^{2}x-1}{3x^{2}}=\dfrac{\tan^{2}x}{3x^{2}}$,仍为 $\tfrac{0}{0}$。(M1) 写成

$$ \frac{\tan^{2}x}{3x^{2}}=\frac{1}{3}\left(\frac{\tan x}{x}\right)^{2}\longrightarrow\frac{1}{3}\cdot 1^{2}=\frac{1}{3}, $$

since $\lim_{x\to 0}\dfrac{\tan x}{x}=1$. (A1)因为 $\lim_{x\to 0}\dfrac{\tan x}{x}=1$。(A1)

Insight.洞察。 Before each application of L'Hopital's rule, re-verify the form is $0/0$ or $\infty/\infty$. The classic trap in (a) and (b) is forgetting to re-check: in (b), each of the three steps is genuinely $\infty/\infty$, so all three applications are valid. In (c), recognising $\tan^{2}x/x^{2}=(\tan x/x)^{2}$ is faster than a second L'Hopital application. The pattern in (b) encodes the growth-rate hierarchy: $e^{x}$ eventually dominates any fixed power of $x$.每次使用洛必达法则前,须重新验证极限形式为 $0/0$ 或 $\infty/\infty$。(a) 和 (b) 中的典型陷阱是忘记重新验证:在 (b) 中,三次使用均为真正的 $\infty/\infty$ 型,故三次均有效。在 (c) 中,识别 $\tan^{2}x/x^{2}=(\tan x/x)^{2}$ 比再次使用洛必达法则更快。(b) 中的模式体现了增长率层级:$e^{x}$ 最终支配 $x$ 的任意固定幂次。
Q4MEDIUMCOREidentifying indeterminate forms and L'Hopital setup识别不定式类型与洛必达法则的应用准备[6 marks]

(a) $\lim_{x\to 0^{+}} x\ln x$; (b) $\lim_{x\to 0}\frac{\sin x}{x+1}$ (state whether L'Hopital applies); (c) $\lim_{x\to\infty}(\sqrt{x^{2}+x}-x)$.(a) $\lim_{x\to 0^{+}} x\ln x$;(b) $\lim_{x\to 0}\frac{\sin x}{x+1}$(说明洛必达法则是否适用);(c) $\lim_{x\to\infty}(\sqrt{x^{2}+x}-x)$。

Answers:答案:  (a) $0$  ·  (b) $0$ (L'Hopital does not apply)(洛必达法则不适用)  ·  (c) $\tfrac{1}{2}$

(a) 0 times infty: rewrite as a quotient(a) $0\cdot\infty$ 型:改写为商 M1·A1

The form is $0\cdot(-\infty)$. Rewrite as $\dfrac{\ln x}{1/x}$, which is $\dfrac{-\infty}{\infty}$. (M1) Apply L'Hopital:极限形式为 $0\cdot(-\infty)$。改写为 $\dfrac{\ln x}{1/x}$,即 $\dfrac{-\infty}{\infty}$ 型。(M1) 使用洛必达法则:

$$ \lim_{x\to 0^{+}}\frac{\ln x}{1/x}=\lim_{x\to 0^{+}}\frac{1/x}{-1/x^{2}}=\lim_{x\to 0^{+}}(-x)=0. $$

(A1)

(b) NOT indeterminate: L'Hopital does not apply(b) 非不定式:洛必达法则不适用 M1·A1

As $x\to 0$: the numerator $\sin x\to 0$ but the denominator $x+1\to 1\ne 0$. (M1) The form is $0/1$, which is not indeterminate. L'Hopital's rule requires the form to be $0/0$ or $\pm\infty/\pm\infty$; neither holds here.当 $x\to 0$ 时:分子 $\sin x\to 0$,但分母 $x+1\to 1\ne 0$。(M1) 形式为 $0/1$,不是不定式。洛必达法则要求形式为 $0/0$ 或 $\pm\infty/\pm\infty$,两者在此均不成立。

$$ \lim_{x\to 0}\frac{\sin x}{x+1}=\frac{\sin 0}{0+1}=\frac{0}{1}=0. $$

(A1)

(c) infty minus infty: rationalise to a quotient(c) $\infty-\infty$ 型:有理化为商 M1·A1

The form is $\infty-\infty$. Multiply by the conjugate: (M1)极限形式为 $\infty-\infty$。乘以共轭表达式:(M1)

$$ \sqrt{x^{2}+x}-x=\frac{(x^{2}+x)-x^{2}}{\sqrt{x^{2}+x}+x}=\frac{x}{\sqrt{x^{2}+x}+x}=\frac{1}{\sqrt{1+1/x}+1}. $$

As $x\to\infty$, $1/x\to 0$, so the expression tends to $\dfrac{1}{\sqrt{1}+1}=\dfrac{1}{2}$. (A1)当 $x\to\infty$ 时,$1/x\to 0$,表达式趋向 $\dfrac{1}{\sqrt{1}+1}=\dfrac{1}{2}$。(A1)

Insight.洞察。 The most dangerous misuse of L'Hopital's rule is applying it when the form is not $0/0$ or $\infty/\infty$. In (b), a student who blindly differentiates gets $\cos x / 1\to 1$, which is wrong. The correct limit is $0$, obtained by direct substitution. L'Hopital is a tool of last resort, not a default: always check whether direct substitution or algebra resolves the limit first.洛必达法则最危险的误用是在极限形式不是 $0/0$ 或 $\infty/\infty$ 时强行使用。在 (b) 中,盲目求导会得到 $\cos x / 1\to 1$,这是错误的。正确极限为 $0$,直接代入即可得到。洛必达法则是最后手段,而非默认工具:始终先检查直接代入或代数化简能否解决极限问题。
PART II  ·  DEFINITIONS AND PROOF第二部分  ·  定义与证明Rigorous arguments · 26 marks严格论证 · 26 分

Worked Solutions详细解答

Q5HARDPROOFstatement of L'Hopital conditions and a caution example洛必达法则条件的完整表述与警示例题[8 marks]

(a) State L'Hopital's rule in full for the $0/0$ form. (b) Explain why L'Hopital cannot be applied to $\lim_{x\to 0}\frac{x^{2}\sin(1/x)}{\sin x}$ and evaluate it correctly.(a) 完整表述 $0/0$ 型洛必达法则。(b) 解释为何洛必达法则不能用于 $\lim_{x\to 0}\frac{x^{2}\sin(1/x)}{\sin x}$,并正确求值。

Answers:答案:  (a) see full statement below见下方完整表述  ·  (b) $0$ (limit of derivative of numerator does not exist; use squeeze instead)(分子导数的极限不存在,应使用夹逼定理)

(a) Full statement of L'Hopital's rule(a) 洛必达法则的完整表述 R1·R1·A1

Hypotheses.假设条件。 Suppose $f$ and $g$ are differentiable on an open interval containing $a$ (except possibly at $a$ itself), that $g'(x)\ne 0$ for $x$ near $a$ but $x\ne a$, and that $\displaystyle\lim_{x\to a}f(x)=0$ and $\displaystyle\lim_{x\to a}g(x)=0$. (R1)设 $f$ 和 $g$ 在含 $a$ 的开区间上可微($a$ 本身可除外),$g'(x)\ne 0$ 在 $a$ 附近(但 $x\ne a$)成立,且 $\displaystyle\lim_{x\to a}f(x)=0$,$\displaystyle\lim_{x\to a}g(x)=0$。(R1)

Suppose further that the limit $\displaystyle\lim_{x\to a}\frac{f'(x)}{g'(x)}$ exists (or equals $\pm\infty$). (R1)进一步设极限 $\displaystyle\lim_{x\to a}\frac{f'(x)}{g'(x)}$ 存在(或等于 $\pm\infty$)。(R1)

Conclusion.结论。 Then $\displaystyle\lim_{x\to a}\frac{f(x)}{g(x)}=\lim_{x\to a}\frac{f'(x)}{g'(x)}$. (A1)则 $\displaystyle\lim_{x\to a}\frac{f(x)}{g(x)}=\lim_{x\to a}\frac{f'(x)}{g'(x)}$。(A1)

(b) Why the application is invalid, and correct evaluation(b) 为何该应用无效,以及正确求值 R1·M1·A1·M1·A1

The form as $x\to 0$ is indeed $0/0$. However, to apply L'Hopital, we need $\displaystyle\lim_{x\to 0}\frac{[x^{2}\sin(1/x)]'}{[\sin x]'}$ to exist. (R1)当 $x\to 0$ 时,极限形式确为 $0/0$。然而,使用洛必达法则需要 $\displaystyle\lim_{x\to 0}\frac{[x^{2}\sin(1/x)]'}{[\sin x]'}$ 存在。(R1)

Compute the derivative of the numerator: $[x^{2}\sin(1/x)]'=2x\sin(1/x)+x^{2}\cos(1/x)\cdot(-1/x^{2})=2x\sin(1/x)-\cos(1/x)$. (M1)计算分子的导数:$[x^{2}\sin(1/x)]'=2x\sin(1/x)+x^{2}\cos(1/x)\cdot(-1/x^{2})=2x\sin(1/x)-\cos(1/x)$。(M1)

As $x\to 0$, the term $\cos(1/x)$ oscillates between $-1$ and $1$ without settling to any limit. Therefore $\displaystyle\lim_{x\to 0}[x^{2}\sin(1/x)]'$ does not exist, and the hypothesis of L'Hopital's rule fails. The rule cannot be applied. (A1)当 $x\to 0$ 时,$\cos(1/x)$ 在 $-1$ 和 $1$ 之间振荡,不趋于任何极限。因此 $\displaystyle\lim_{x\to 0}[x^{2}\sin(1/x)]'$ 不存在,洛必达法则的假设条件不满足,该法则不可使用。(A1)

Correct method.正确方法。 Write the limit as a product and use the bound $|\sin(1/x)|\le 1$: (M1)将极限写为乘积,利用界 $|\sin(1/x)|\le 1$:(M1)

$$ \frac{x^{2}\sin(1/x)}{\sin x}=\underbrace{\frac{x}{\sin x}}_{\to\, 1}\cdot\underbrace{x\sin(1/x)}_{\to\, 0}, $$

since $x\sin(1/x)\to 0$ by the squeeze theorem ($|x\sin(1/x)|\le|x|\to 0$) and $x/\sin x\to 1$. The limit is $1\cdot 0=0$. (A1)因为由夹逼定理($|x\sin(1/x)|\le|x|\to 0$)得 $x\sin(1/x)\to 0$,且 $x/\sin x\to 1$。极限为 $1\cdot 0=0$。(A1)

Insight.洞察。 L'Hopital's rule has a hidden hypothesis that many textbooks underemphasise: the limit $\lim f'/g'$ must itself exist. If $f'(x)/g'(x)$ oscillates or diverges, the rule is simply silent. The failure mode here is a derivative that does not exist as a limit, even though the original quotient does. Whenever the numerator contains a wildly oscillating factor, the squeeze theorem is the correct tool.洛必达法则有一个许多教材未充分强调的隐含条件:极限 $\lim f'/g'$ 本身必须存在。若 $f'(x)/g'(x)$ 振荡或发散,该法则无能为力。此处的失效原因是导数作为极限不存在,尽管原始商的极限存在。当分子含有剧烈振荡因子时,夹逼定理才是正确工具。
Q6HARDPROOFexponential indeterminate forms via logarithms利用对数处理指数型不定式[10 marks]

Evaluate using the logarithm technique: (a) $\lim_{x\to 0^{+}}x^{x}$ (form $0^{0}$); (b) $\lim_{x\to\infty}x^{1/x}$ (form $\infty^{0}$); (c) $\lim_{x\to\infty}(1+3/x)^{x}$ (form $1^{\infty}$).用对数技巧求值:(a) $\lim_{x\to 0^{+}}x^{x}$($0^{0}$ 型);(b) $\lim_{x\to\infty}x^{1/x}$($\infty^{0}$ 型);(c) $\lim_{x\to\infty}(1+3/x)^{x}$($1^{\infty}$ 型)。

Answers:答案:  (a) $1$  ·  (b) $1$  ·  (c) $e^{3}$

(a) x^x as x approaches 0+: form 0^0(a) $x\to 0^+$ 时 $x^x$:$0^0$ 型 M1·A1·A1

Let $y=x^{x}$, so $\ln y=x\ln x$. This is the $0\cdot(-\infty)$ form. Rewrite as $\dfrac{\ln x}{1/x}$. (M1)令 $y=x^{x}$,则 $\ln y=x\ln x$,为 $0\cdot(-\infty)$ 型。改写为 $\dfrac{\ln x}{1/x}$。(M1)

$$ \lim_{x\to 0^{+}}\frac{\ln x}{1/x}\overset{L'H}{=}\lim_{x\to 0^{+}}\frac{1/x}{-1/x^{2}}=\lim_{x\to 0^{+}}(-x)=0. $$

Therefore $\lim_{x\to 0^{+}}\ln y=0$, so $\lim_{x\to 0^{+}}y=e^{0}=1$. (A1·A1)故 $\lim_{x\to 0^{+}}\ln y=0$,从而 $\lim_{x\to 0^{+}}y=e^{0}=1$。(A1·A1)

(b) x^{1/x} as x approaches infty: form infty^0(b) $x\to\infty$ 时 $x^{1/x}$:$\infty^0$ 型 M1·A1·A1

Let $y=x^{1/x}$, so $\ln y=\dfrac{\ln x}{x}$. This is $\infty/\infty$. (M1)令 $y=x^{1/x}$,则 $\ln y=\dfrac{\ln x}{x}$,为 $\infty/\infty$ 型。(M1)

$$ \lim_{x\to\infty}\frac{\ln x}{x}\overset{L'H}{=}\lim_{x\to\infty}\frac{1/x}{1}=\lim_{x\to\infty}\frac{1}{x}=0. $$

Therefore $\lim_{x\to\infty}\ln y=0$, so $\lim_{x\to\infty}y=e^{0}=1$. (A1·A1)故 $\lim_{x\to\infty}\ln y=0$,从而 $\lim_{x\to\infty}y=e^{0}=1$。(A1·A1)

(c) (1 + 3/x)^x as x approaches infty: form 1^infty(c) $x\to\infty$ 时 $(1+3/x)^x$:$1^\infty$ 型 M1·M1·A1·A1

Let $y=\left(1+\dfrac{3}{x}\right)^{x}$, so $\ln y=x\ln\!\left(1+\dfrac{3}{x}\right)$. This is $\infty\cdot 0$. Rewrite: (M1)令 $y=\left(1+\dfrac{3}{x}\right)^{x}$,则 $\ln y=x\ln\!\left(1+\dfrac{3}{x}\right)$,为 $\infty\cdot 0$ 型。改写:(M1)

$$ \ln y=\frac{\ln(1+3/x)}{1/x}. $$

The form is $0/0$ as $x\to\infty$. Apply L'Hopital: (M1)当 $x\to\infty$ 时形式为 $0/0$。使用洛必达法则:(M1)

$$ \lim_{x\to\infty}\frac{\ln(1+3/x)}{1/x}\overset{L'H}{=}\lim_{x\to\infty}\frac{\dfrac{-3/x^{2}}{1+3/x}}{-1/x^{2}}=\lim_{x\to\infty}\frac{3}{1+3/x}=3. $$

Therefore $\lim_{x\to\infty}\ln y=3$, so $\lim_{x\to\infty}y=e^{3}$. (A1·A1)故 $\lim_{x\to\infty}\ln y=3$,从而 $\lim_{x\to\infty}y=e^{3}$。(A1·A1)

Insight.洞察。 The logarithm technique converts every exponential indeterminate form $u^{v}$ into a product $v\ln u$, which is then $0\cdot\pm\infty$ and can be rewritten as a $0/0$ or $\infty/\infty$ quotient for L'Hopital. The chain is always: set $y=u^{v}$, take $\ln$, evaluate $\lim\ln y$, recover $\lim y=e^{\lim\ln y}$. Part (c) is the prototype of the compound-interest limit: the general result $\lim_{x\to\infty}(1+a/x)^{x}=e^{a}$ follows by the same steps with $3$ replaced by $a$.对数技巧将所有指数型不定式 $u^{v}$ 转化为乘积 $v\ln u$(即 $0\cdot\pm\infty$ 型),进而改写为 $0/0$ 或 $\infty/\infty$ 型商,以便使用洛必达法则。推导链始终是:令 $y=u^{v}$,取对数,求 $\lim\ln y$,还原 $\lim y=e^{\lim\ln y}$。(c) 是复利极限的原型:将 $3$ 替换为 $a$,可用相同步骤得到一般结论 $\lim_{x\to\infty}(1+a/x)^{x}=e^{a}$。
Q7HARDPROOFNewton's method iteration and convergence condition牛顿迭代法与收敛条件[8 marks]

Newton's method for $f(x)=x^{2}-3$: (a) two iterations from $x_{0}=1$; (b) geometric meaning of each step; (c) one condition that causes failure.对 $f(x)=x^{2}-3$ 使用牛顿迭代法:(a) 从 $x_{0}=1$ 出发进行两次迭代;(b) 每步的几何含义;(c) 导致失效的一个条件。

Answers:答案:  (a) $x_{1}=2.0000$, $x_{2}=1.7500$  ·  (b) $x$-intercept of the tangent line at $(x_{n},f(x_{n}))$切线在 $(x_{n},f(x_{n}))$ 处与 $x$ 轴的交点  ·  (c) $f'(x_{n})\approx 0$ causes divergence$f'(x_{n})\approx 0$ 导致发散

(a) Two Newton iterations(a) 两次牛顿迭代 M1·A1·M1·A1

$f(x)=x^{2}-3$, $f'(x)=2x$. The iteration is $x_{n+1}=x_{n}-\dfrac{x_{n}^{2}-3}{2x_{n}}$. (M1)$f(x)=x^{2}-3$,$f'(x)=2x$。迭代公式为 $x_{n+1}=x_{n}-\dfrac{x_{n}^{2}-3}{2x_{n}}$。(M1)

$x_{1}=1-\dfrac{1-3}{2}=1+1=2.0000$. (A1)$x_{1}=1-\dfrac{1-3}{2}=1+1=2.0000$。(A1)

(M1) $x_{2}=2-\dfrac{4-3}{4}=2-0.25=1.7500$. (A1) (True value: $\sqrt{3}\approx 1.7321$.)(M1) $x_{2}=2-\dfrac{4-3}{4}=2-0.25=1.7500$。(A1)(真实值:$\sqrt{3}\approx 1.7321$。)

(b) Geometric interpretation(b) 几何解释 R1·A1

At each step, draw the tangent line to $y=f(x)$ at the current point $(x_{n},f(x_{n}))$. (R1) The next iterate $x_{n+1}$ is the $x$-intercept of that tangent line: solving $0=f(x_{n})+f'(x_{n})(x-x_{n})$ gives $x=x_{n}-f(x_{n})/f'(x_{n})$. (A1)每步在当前点 $(x_{n},f(x_{n}))$ 处作 $y=f(x)$ 的切线。(R1) 下一个迭代值 $x_{n+1}$ 是该切线与 $x$ 轴的交点:解方程 $0=f(x_{n})+f'(x_{n})(x-x_{n})$ 得 $x=x_{n}-f(x_{n})/f'(x_{n})$。(A1)

(c) Condition for failure(c) 失效条件 R1·A1

If $f'(x_{n})\approx 0$ (the function is nearly flat at the current iterate), the tangent line is nearly horizontal and its $x$-intercept is enormously far from $x_{n}$. (R1) For example, applying Newton's method to $f(x)=x^{3}-x$ starting near the inflection point $x=0$ can send the iterate far from any root. (A1)若 $f'(x_{n})\approx 0$(函数在当前迭代点附近近乎水平),则切线几乎水平,其与 $x$ 轴的交点距 $x_{n}$ 极远。(R1) 例如,对 $f(x)=x^{3}-x$ 从接近拐点 $x=0$ 处出发使用牛顿迭代法,迭代值可能远离任何根。(A1)

Insight.洞察。 Newton's method achieves quadratic convergence when it works: the number of correct decimal digits roughly doubles with each step (visible in (a): from no digits to 1 to 2). This speed comes at the cost of needing a good starting guess; a poor $x_{0}$ near a flat region or near a local extremum of $f$ can cause the method to diverge or cycle between points. The linearization from Section 1 of the SG is exactly what Newton's method exploits at every step.牛顿迭代法在有效时可实现二次收敛:正确小数位数大约每步翻倍((a) 中可见:从 0 位到 1 位再到 2 位)。这种速度以需要一个好的初始猜测值为代价;初始值 $x_{0}$ 若处于函数近乎水平的区域或局部极值附近,可能导致方法发散或在点间循环。学习指南第 1 节中的线性化正是牛顿迭代法每步所利用的工具。
PART III  ·  APPLICATIONS AND SYNTHESIS第三部分  ·  应用与综合Extended problems · 28 marks综合拓展题 · 28 分

Worked Solutions详细解答

Q8HARDAPPLIEDrepeated L'Hopital and a looping caution多次洛必达与循环陷阱警示[8 marks]

(a) $\lim_{x\to 0}\frac{e^{x}-1-x-\tfrac{1}{2}x^{2}}{x^{3}}$; (b) $\lim_{x\to 0}\frac{1-\cos x}{x^{2}}$ via two L'Hopital steps; (c) explain and resolve the loop in $\lim_{x\to\infty}\frac{e^{x}+e^{-x}}{e^{x}-e^{-x}}$.(a) $\lim_{x\to 0}\frac{e^{x}-1-x-\tfrac{1}{2}x^{2}}{x^{3}}$;(b) $\lim_{x\to 0}\frac{1-\cos x}{x^{2}}$(两步洛必达法则);(c) 解释并解决 $\lim_{x\to\infty}\frac{e^{x}+e^{-x}}{e^{x}-e^{-x}}$ 中的循环问题。

Answers:答案:  (a) $\tfrac{1}{6}$  ·  (b) $\tfrac{1}{2}$  ·  (c) $1$ (L'Hopital loops; divide by $e^{x}$)(洛必达循环,除以 $e^{x}$)

(a) Three applications of L'Hopital(a) 三次使用洛必达法则 M1·M1·M1·A1

At $x=0$, numerator and denominator are both $0$ (form $0/0$). (M1) First application:在 $x=0$ 处,分子和分母均为 $0$($0/0$ 型)。(M1) 第一次使用:

$$ \frac{e^{x}-1-x}{3x^{2}}\quad\text{still }0/0\text{ at }x=0. $$

(M1) Second application:(M1) 第二次使用:

$$ \frac{e^{x}-1}{6x}\quad\text{still }0/0\text{ at }x=0. $$

(M1) Third application:(M1) 第三次使用:

$$ \frac{e^{x}}{6}\longrightarrow\frac{e^{0}}{6}=\frac{1}{6}. $$

(A1)

(b) Two L'Hopital applications: re-verify at each step(b) 两次洛必达法则:每步重新验证 M1·M1·A1

Form is $0/0$ at $x=0$. (M1) First step: $\dfrac{\sin x}{2x}$. This is still $0/0$ at $x=0$. (M1) Second step:在 $x=0$ 处为 $0/0$ 型。(M1) 第一步:$\dfrac{\sin x}{2x}$,在 $x=0$ 处仍为 $0/0$。(M1) 第二步:

$$ \frac{\cos x}{2}\longrightarrow\frac{1}{2}. $$

(A1)

(c) L'Hopital loops: use direct algebra instead(c) 洛必达法则陷入循环:改用直接代数化简 R1·M1·A1

The form as $x\to\infty$ is $\infty/\infty$. Differentiating: numerator becomes $e^{x}-e^{-x}$, denominator $e^{x}+e^{-x}$. This is the original quotient with numerator and denominator swapped, so applying L'Hopital again returns to $(e^{x}+e^{-x})/(e^{x}-e^{-x})$. The rule loops indefinitely without resolving the limit. (R1)当 $x\to\infty$ 时形式为 $\infty/\infty$。求导后:分子变为 $e^{x}-e^{-x}$,分母为 $e^{x}+e^{-x}$。这是原商的分子分母互换,再次使用洛必达法则将回到 $(e^{x}+e^{-x})/(e^{x}-e^{-x})$。法则无限循环,无法求出极限。(R1)

Instead, divide numerator and denominator by $e^{x}$: (M1)改为将分子分母均除以 $e^{x}$:(M1)

$$ \frac{e^{x}+e^{-x}}{e^{x}-e^{-x}}=\frac{1+e^{-2x}}{1-e^{-2x}}\longrightarrow\frac{1+0}{1-0}=1. $$

(A1)

Insight.洞察。 L'Hopital's rule can loop when numerator and denominator are related by differentiation in a circular way, as here with $\cosh x/\sinh x$. The loop is not a sign of an invalid application; the hypotheses are satisfied each time. But the rule produces no information in a cycle. The remedy is always to look for algebraic simplification first: dividing by the dominant exponential resolves the limit instantly and reveals that the limit equals $1$ (i.e. $\tanh x\to 1$ as $x\to\infty$).当分子和分母通过求导以循环方式关联时,洛必达法则会陷入循环,如此处的 $\cosh x/\sinh x$。循环并非无效使用的标志,每次均满足假设条件。但法则在循环中不产生任何信息。解决方案始终是先寻找代数化简:除以主导指数项立即解决极限,并揭示极限等于 $1$(即 $\tanh x\to 1$,当 $x\to\infty$)。
Q9HARDAPPLIEDgrowth rates, 0 times infty, and infty minus infty增长率、$0\cdot\infty$ 型与 $\infty-\infty$ 型[10 marks]

(a) Prove $\lim_{x\to\infty}x^{n}/e^{x}=0$ for all positive integers $n$; (b) evaluate $\lim_{x\to 0^{+}}(1/\sin x - 1/x)$; (c) evaluate $\lim_{x\to\infty}(\ln(x+1)-\ln x)$ without L'Hopital.(a) 证明对所有正整数 $n$,$\lim_{x\to\infty}x^{n}/e^{x}=0$;(b) 求 $\lim_{x\to 0^{+}}(1/\sin x - 1/x)$;(c) 不用洛必达法则求 $\lim_{x\to\infty}(\ln(x+1)-\ln x)$。

Answers:答案:  (a) $0$ for all $n$对所有 $n$  ·  (b) $0$  ·  (c) $0$

(a) Dominance of e^x over every polynomial(a) $e^x$ 对所有多项式的支配性 M1·M1·A1·A1

The form is $\infty/\infty$. Apply L'Hopital once to reduce the power of $x$: (M1)极限形式为 $\infty/\infty$。使用一次洛必达法则以降低 $x$ 的幂次:(M1)

$$ \frac{x^{n}}{e^{x}}\overset{L'H}{\longrightarrow}\frac{nx^{n-1}}{e^{x}}\overset{L'H}{\longrightarrow}\frac{n(n-1)x^{n-2}}{e^{x}}\overset{L'H}{\longrightarrow}\cdots $$

Each application reduces the exponent of $x$ by $1$. The form remains $\infty/\infty$ as long as the exponent in the numerator is positive. (M1) After exactly $n$ applications, the numerator becomes the constant $n!$:每次使用将 $x$ 的指数降低 $1$。只要分子中的指数为正,形式始终保持 $\infty/\infty$。(M1) 恰好 $n$ 次使用后,分子变为常数 $n!$:

$$ \lim_{x\to\infty}\frac{n!}{e^{x}}=0, $$

since $e^{x}\to\infty$. (A1) Therefore $\lim_{x\to\infty}x^{n}/e^{x}=0$ for every positive integer $n$. (A1)因为 $e^{x}\to\infty$。(A1) 故对每个正整数 $n$,$\lim_{x\to\infty}x^{n}/e^{x}=0$。(A1)

(b) infty minus infty: combine over a common denominator(b) $\infty-\infty$ 型:通分 M1·M1·A1

Combine the two fractions: (M1)将两个分数合并:(M1)

$$ \frac{1}{\sin x}-\frac{1}{x}=\frac{x-\sin x}{x\sin x}. $$

As $x\to 0^{+}$, both numerator and denominator tend to $0$ (form $0/0$). Apply L'Hopital: (M1)当 $x\to 0^{+}$ 时,分子和分母均趋向 $0$($0/0$ 型)。使用洛必达法则:(M1)

$$ \frac{1-\cos x}{\sin x+x\cos x}\overset{L'H}{=}\frac{\sin x}{2\cos x-x\sin x}\longrightarrow\frac{0}{2-0}=0. $$

(A1)

(c) Logarithm law avoids L'Hopital(c) 利用对数法则,无需洛必达法则 M1·A1·R1

Use the logarithm subtraction rule: (M1)利用对数相减法则:(M1)

$$ \ln(x+1)-\ln x=\ln\!\left(\frac{x+1}{x}\right)=\ln\!\left(1+\frac{1}{x}\right). $$

As $x\to\infty$, $1/x\to 0$, so $\ln(1+1/x)\to\ln 1=0$. (A1)当 $x\to\infty$ 时,$1/x\to 0$,故 $\ln(1+1/x)\to\ln 1=0$。(A1)

This is consistent with the growth-rate hierarchy: $\ln x$ grows so slowly that the difference between consecutive values $\ln(x+1)$ and $\ln x$ shrinks to zero, even though each value diverges. (R1)这与增长率层级一致:$\ln x$ 增长极为缓慢,以至于相邻值 $\ln(x+1)$ 与 $\ln x$ 之差趋向零,尽管每个值本身趋向无穷。(R1)

Insight.洞察。 Part (a) is the quantitative statement of the growth-rate hierarchy: polynomial beats logarithm, exponential beats polynomial. The proof by repeated L'Hopital is the standard argument: each step peels off one factor of $x$ from the numerator while $e^{x}$ is unchanged, and after $n$ steps the numerator is a constant. Part (b) illustrates the standard handling of $\infty-\infty$: always combine into a single fraction before reaching for L'Hopital.(a) 是增长率层级的定量表述:多项式胜过对数,指数胜过多项式。反复使用洛必达法则的证明是标准论证:每步从分子剥去一个 $x$ 因子,而 $e^{x}$ 保持不变,经过 $n$ 步后分子为常数。(b) 展示了处理 $\infty-\infty$ 型的标准方法:在使用洛必达法则之前,始终先通分。
Q10HARDAPPLIEDlinearization error bound and the natural exponential limit线性化误差界与自然指数极限[10 marks]

(a) Evaluate $\lim_{x\to 0^{+}}(1+x)^{1/x}$; (b) use the linearization of $\ln(1+t)$ at $t=0$ to explain why the answer is $e$; (c) compare the absolute errors of $L(x)=1+x$ and $Q(x)=1+x+\tfrac{1}{2}x^{2}$ as approximations to $e^{x}$ at $x=0.2$.(a) 求 $\lim_{x\to 0^{+}}(1+x)^{1/x}$;(b) 利用 $\ln(1+t)$ 在 $t=0$ 处的线性化解释结果为何等于 $e$;(c) 比较 $L(x)=1+x$ 与 $Q(x)=1+x+\tfrac{1}{2}x^{2}$ 在 $x=0.2$ 处近似 $e^{x}$ 的绝对误差。

Answers:答案:  (a) $e$  ·  (b) the linearization $\ln(1+t)\approx t$ makes the exponent approach $1$线性化 $\ln(1+t)\approx t$ 使指数趋向 $1$  ·  (c) linear error $\approx 0.0214$; quadratic error $\approx 0.0014$; $Q$ is roughly $15\times$ more accurate线性误差约 $0.0214$;二次误差约 $0.0014$;$Q$ 的精度约为 $L$ 的 $15$ 倍

(a) Limit of (1+x)^{1/x} as x approaches 0+: logarithm technique(a) $x\to 0^+$ 时 $(1+x)^{1/x}$ 的极限:对数技巧 M1·M1·A1·A1

Let $y=(1+x)^{1/x}$, so $\ln y=\dfrac{\ln(1+x)}{x}$. (M1) As $x\to 0^{+}$, both numerator and denominator tend to $0$ (form $0/0$). Apply L'Hopital: (M1)令 $y=(1+x)^{1/x}$,则 $\ln y=\dfrac{\ln(1+x)}{x}$。(M1) 当 $x\to 0^{+}$ 时,分子分母均趋向 $0$($0/0$ 型)。使用洛必达法则:(M1)

$$ \lim_{x\to 0^{+}}\frac{\ln(1+x)}{x}\overset{L'H}{=}\lim_{x\to 0^{+}}\frac{1/(1+x)}{1}=\frac{1}{1+0}=1. $$

Therefore $\lim_{x\to 0^{+}}\ln y=1$, so $\lim_{x\to 0^{+}}y=e^{1}=e$. (A1·A1)故 $\lim_{x\to 0^{+}}\ln y=1$,从而 $\lim_{x\to 0^{+}}y=e^{1}=e$。(A1·A1)

(b) Intuitive explanation via linearization of ln(b) 通过对数的线性化给出直观解释 M1·A1·R1

Linearize $g(t)=\ln(1+t)$ at $t=0$: $g'(t)=\dfrac{1}{1+t}$, so $g'(0)=1$. (M1) The linearization is $L(t)=0+1\cdot t=t$, giving the approximation $\ln(1+t)\approx t$ for small $t$. (A1)将 $g(t)=\ln(1+t)$ 在 $t=0$ 处线性化:$g'(t)=\dfrac{1}{1+t}$,故 $g'(0)=1$。(M1) 线性化为 $L(t)=0+1\cdot t=t$,给出小 $t$ 时的近似 $\ln(1+t)\approx t$。(A1)

Applying this with $t=x$ small:将此应用于小 $t=x$:

$$ \ln y=\frac{\ln(1+x)}{x}\approx\frac{x}{x}=1. $$

Therefore $y\approx e^{1}=e$. This gives an intuitive understanding of why $e$ is defined by this limit: as $x\to 0^{+}$, the linearization of $\ln$ makes the exponent of $e$ approach exactly $1$. (R1)故 $y\approx e^{1}=e$。这直观地解释了为何 $e$ 由此极限定义:当 $x\to 0^{+}$ 时,对数的线性化使 $e$ 的指数恰好趋向 $1$。(R1)

(c) Comparing linear and quadratic error at x = 0.2(c) 在 $x=0.2$ 处比较线性与二次误差 M1·A1·A1

Use $e^{0.2}\approx 1.2214$ (given). (M1)使用 $e^{0.2}\approx 1.2214$(已知)。(M1)

Linear approximation: $L(0.2)=1+0.2=1.2000$. Absolute error: $|1.2214-1.2000|=0.0214$. (A1)线性近似:$L(0.2)=1+0.2=1.2000$。绝对误差:$|1.2214-1.2000|=0.0214$。(A1)

Quadratic approximation: $Q(0.2)=1+0.2+\tfrac{1}{2}(0.04)=1+0.2+0.02=1.2200$. Absolute error: $|1.2214-1.2200|=0.0014$. (A1)二次近似:$Q(0.2)=1+0.2+\tfrac{1}{2}(0.04)=1+0.2+0.02=1.2200$。绝对误差:$|1.2214-1.2200|=0.0014$。(A1)

The ratio of errors is $0.0214/0.0014\approx 15.3$, so the quadratic approximation is roughly $15$ times more accurate at $x=0.2$.误差之比为 $0.0214/0.0014\approx 15.3$,故在 $x=0.2$ 处,二次近似的精度约为线性近似的 $15$ 倍。

Insight.洞察。 Part (a) is a classic limit that defines $e$ from a different angle: $e=\lim_{x\to 0}(1+x)^{1/x}=\lim_{n\to\infty}(1+1/n)^{n}$. Parts (b) and (c) illustrate why higher-order approximations are worth the extra term: adding the quadratic term $\tfrac{1}{2}x^{2}$ reduced the error by a factor of about $15$ at $x=0.2$. In general, the $n$-th order Taylor polynomial reduces the error by a factor of roughly $x$ at each additional term, so for $x=0.2$ each extra term buys about a factor of $5$ in accuracy, and the quadratic offers a further factor of $3$ from the specific coefficient $\tfrac{1}{2}$.(a) 是从另一角度定义 $e$ 的经典极限:$e=\lim_{x\to 0}(1+x)^{1/x}=\lim_{n\to\infty}(1+1/n)^{n}$。(b) 和 (c) 说明高阶近似为何值得额外的项:加入二次项 $\tfrac{1}{2}x^{2}$ 在 $x=0.2$ 处将误差缩小了约 $15$ 倍。一般来说,$n$ 阶泰勒多项式每增加一项约将误差缩小 $x$ 倍,因此对于 $x=0.2$,每个额外项提高约 $5$ 倍精度,而二次项还因系数 $\tfrac{1}{2}$ 额外提供了约 $3$ 倍的精度提升。