University Calculus · Calculus I大学微积分 · 微积分 I

Unit A7: Curve Sketching and Optimization单元 A7:曲线描绘与最优化

From critical points and the Mean Value Theorem to the sign of the first and second derivatives, this unit turns differentiation into a toolkit for reading and optimizing the shape of a graph.从临界点(critical point)、中值定理(Mean Value Theorem),到一阶、二阶导数的正负号,本单元把求导变成一套读懂并优化图像形状的工具。

Calculus I微积分 I Single-Variable单变量 Foundational基础 MIT 18.01 / GT 1551 / Princeton MAT 103MIT 18.01 / GT 1551 / Princeton MAT 103
Read me first. These notes assume you can already differentiate fluently and want to use derivatives to understand behavior, not just compute slopes. Work the seven sections in order: each builds the candidate list (critical points), justifies the sign analysis (the Mean Value Theorem), and then applies it to monotonicity, concavity, full curve sketching, and both pure and applied optimization. Open every worked example and try it before reading the solution, then test yourself on the unit quiz.
先读这里。 本讲义假定你已经能熟练求导,目标是用导数(derivative)理解函数的行为,而不只是算斜率。请按顺序学习七节:每一节先建立候选名单(临界点 critical point),再为正负号分析提供依据(中值定理 Mean Value Theorem),然后应用到单调性、凹凸性、完整的曲线描绘,以及纯粹与应用两类最优化(optimization)。每个例题都先自己动手再看解答,最后用单元测验检验自己。

Extreme Values and Critical Points极值与临界点

Key idea. A function attains a global maximum or minimum only at a critical point or at an endpoint of its domain. Critical points are where the first derivative vanishes or fails to exist, so they form the short list of candidates that organizes every optimization and curve-sketching argument.
核心思想。函数只可能在临界点(critical point)或定义域端点处取得全局最大值或最小值。临界点是一阶导数(derivative)为零或不存在之处,因此它们构成一份简短的候选名单,统领所有最优化与曲线描绘的论证。

Definition. Let $f$ be defined on a set $D$ and let $c \in D$. We say $f$ has an absolute (global) maximum at $c$ if $f(c) \ge f(x)$ for all $x \in D$, and an absolute minimum at $c$ if $f(c) \le f(x)$ for all $x \in D$. We say $f$ has a local (relative) maximum at $c$ if $f(c) \ge f(x)$ for all $x$ in some open interval containing $c$, with the analogous definition for a local minimum.

定义。设 $f$ 定义在集合 $D$ 上,$c \in D$。若对一切 $x \in D$ 都有 $f(c) \ge f(x)$,则称 $f$ 在 $c$ 处取得绝对(全局)最大值absolute maximum);若对一切 $x \in D$ 都有 $f(c) \le f(x)$,则取得绝对最小值。若仅对某个含 $c$ 的开区间内的所有 $x$ 有 $f(c) \ge f(x)$,则称 $f$ 在 $c$ 处取得局部(相对)最大值local maximum),局部最小值类似定义。

Definition. A number $c$ in the domain of $f$ is a critical point of $f$ if either $f'(c) = 0$ or $f'(c)$ does not exist.

定义。若 $f$ 定义域中的数 $c$ 满足 $f'(c) = 0$ 或 $f'(c)$ 不存在,则称 $c$ 为 $f$ 的临界点critical point)。

Fermat's Theorem费马定理(Fermat's Theorem)
$$ f \text{ has a local extremum at } c \text{ and } f'(c) \text{ exists} \implies f'(c) = 0. $$

Theorem (Extreme Value Theorem). If $f$ is continuous on a closed bounded interval $[a,b]$, then $f$ attains an absolute maximum value and an absolute minimum value on $[a,b]$. Continuity and the closed interval are both essential: $f(x) = 1/x$ on $(0,1]$ is continuous but unbounded, and so attains no maximum.

定理(极值定理,Extreme Value Theorem / EVT)。若 $f$ 在有界闭区间 $[a,b]$ 上连续(continuous),则 $f$ 在 $[a,b]$ 上必取得绝对最大值与绝对最小值。连续性与闭区间二者缺一不可:$f(x) = 1/x$ 在 $(0,1]$ 上连续却无界,因而取不到最大值。

Remark. The converse of Fermat's Theorem is false. For $f(x) = x^3$ we have $f'(0) = 0$, yet $0$ is neither a local maximum nor a local minimum. A critical point is a candidate, not a guarantee.

注。费马定理的逆命题不成立。对 $f(x) = x^3$ 有 $f'(0) = 0$,但 $0$ 既非局部最大值也非局部最小值。临界点只是候选,并不保证是极值。

Worked Example 1.1: locating critical points例题 1.1:求临界点

Find the critical points of $f(x) = x^{3} - 3x^{2} - 9x + 5$.

Differentiate and factor.

$$ f'(x) = 3x^{2} - 6x - 9 = 3(x^{2} - 2x - 3) = 3(x-3)(x+1). $$

Since $f'$ is a polynomial it exists everywhere, so the only critical points come from $f'(x) = 0$, giving $x = 3$ and $x = -1$. These two numbers are the complete candidate list for the local extrema of $f$.

求 $f(x) = x^{3} - 3x^{2} - 9x + 5$ 的临界点。

求导并因式分解。

$$ f'(x) = 3x^{2} - 6x - 9 = 3(x^{2} - 2x - 3) = 3(x-3)(x+1). $$

由于 $f'$ 是多项式,处处存在,所以临界点只能来自 $f'(x) = 0$,即 $x = 3$ 与 $x = -1$。这两个数就是 $f$ 局部极值的完整候选名单。

Worked Example 1.2: a critical point where the derivative is undefined例题 1.2:导数不存在处的临界点

Find the critical points of $f(x) = x^{2/3}$.

$$ f'(x) = \tfrac{2}{3} x^{-1/3} = \frac{2}{3\,x^{1/3}}. $$

This is never zero, but it fails to exist at $x = 0$, where $f$ is still defined. Hence $x = 0$ is a critical point. Inspecting $f$ shows it has an absolute minimum there: this candidate would be invisible to anyone who only solved $f'(x) = 0$.

求 $f(x) = x^{2/3}$ 的临界点。

$$ f'(x) = \tfrac{2}{3} x^{-1/3} = \frac{2}{3\,x^{1/3}}. $$

它永不为零,但在 $x = 0$ 处不存在,而 $f$ 在该点仍有定义。因此 $x = 0$ 是临界点。考察 $f$ 可知它在此取得绝对最小值:只解 $f'(x) = 0$ 的人会完全漏掉这个候选。

Worked Example 1.3: a transcendental candidate list例题 1.3:超越函数的候选名单

Find the critical points of $f(x) = x - 2\sin x$ on the interval $[0, 2\pi]$.

The function is differentiable everywhere, so every critical point comes from $f'(x) = 0$.

$$ f'(x) = 1 - 2\cos x = 0 \implies \cos x = \tfrac{1}{2}. $$

On $[0,2\pi]$ the equation $\cos x = \tfrac12$ has the two solutions $x = \tfrac{\pi}{3}$ and $x = \tfrac{5\pi}{3}$. These are the only interior critical points. When the problem is later treated as an extremum question on the closed interval, the full candidate list will be $\{0,\ \tfrac{\pi}{3},\ \tfrac{5\pi}{3},\ 2\pi\}$: the two critical points plus the two endpoints. The lesson is that solving the trigonometric equation must respect the stated domain, since $\cos x = \tfrac12$ has infinitely many solutions on the whole line.

在区间 $[0, 2\pi]$ 上求 $f(x) = x - 2\sin x$ 的临界点。

该函数处处可微(differentiable),所以每个临界点都来自 $f'(x) = 0$。

$$ f'(x) = 1 - 2\cos x = 0 \implies \cos x = \tfrac{1}{2}. $$

在 $[0,2\pi]$ 上,方程 $\cos x = \tfrac12$ 有两个解 $x = \tfrac{\pi}{3}$ 与 $x = \tfrac{5\pi}{3}$,这是仅有的内部临界点。当此题后续作为闭区间上的极值问题处理时,完整的候选名单为 $\{0,\ \tfrac{\pi}{3},\ \tfrac{5\pi}{3},\ 2\pi\}$:两个临界点加两个端点。要点在于解三角方程必须尊重题目给定的定义域,因为 $\cos x = \tfrac12$ 在整条数轴上有无穷多个解。

Worked Example 1.4: the Extreme Value Theorem can fail on an open interval例题 1.4:极值定理在开区间上可能失效

Show that $f(x) = x$ on the open interval $(0,1)$ has neither an absolute maximum nor an absolute minimum, even though $f$ is continuous and bounded.

Suppose, for contradiction, that $f$ attained a maximum at some $c \in (0,1)$. Then $f(c) = c$, yet the number $\tfrac{c+1}{2}$ also lies in $(0,1)$ and satisfies $\tfrac{c+1}{2} > c$, so $f\big(\tfrac{c+1}{2}\big) > f(c)$, contradicting maximality. The symmetric argument with $\tfrac{c}{2}$ rules out a minimum. The values approach $1$ and $0$ but are never reached. This is exactly why the Extreme Value Theorem insists on a closed interval: the supremum and infimum exist as numbers, but without the endpoints there is no point at which they are attained.

证明 $f(x) = x$ 在开区间 $(0,1)$ 上既无绝对最大值也无绝对最小值,尽管 $f$ 连续且有界。

反证:设 $f$ 在某点 $c \in (0,1)$ 取得最大值。则 $f(c) = c$,但数 $\tfrac{c+1}{2}$ 也属于 $(0,1)$ 且满足 $\tfrac{c+1}{2} > c$,于是 $f\big(\tfrac{c+1}{2}\big) > f(c)$,与最大性矛盾。用 $\tfrac{c}{2}$ 作对称论证可排除最小值。函数值趋于 $1$ 和 $0$,却永远取不到。这正是极值定理坚持要求闭区间的原因:上确界与下确界作为数是存在的,但缺了端点就没有任何一点真正取到它们。

Common error. Many students equate "critical point" with "$f'(x) = 0$" and stop there. The definition has two clauses: $f'(c) = 0$ or $f'(c)$ fails to exist. Functions with corners, cusps, or vertical tangents (such as $|x|$, $x^{2/3}$, and $x^{1/3}$) hide extrema precisely at the points where the derivative blows up or jumps. A second frequent slip is to call a point critical when it is not in the domain at all: for $f(x)=1/x$ the value $x=0$ is not a critical point, because a critical point must belong to the domain of $f$. Always confirm membership in the domain before adding a candidate to the list.
常见错误。许多学生把"临界点"等同于"$f'(x) = 0$"就此打住。定义有两种情形:$f'(c) = 0$ $f'(c)$ 不存在。带尖角、尖点或竖直切线的函数(如 $|x|$、$x^{2/3}$、$x^{1/3}$)恰恰把极值藏在导数爆掉或跳变之处。第二个常见疏漏是把根本不在定义域内的点当成临界点:对 $f(x)=1/x$,$x=0$ 不是临界点,因为临界点必须属于 $f$ 的定义域。把候选加入名单前,务必先确认它在定义域内。
Which statement about the critical points of $f(x) = |x|$ is correct?关于 $f(x) = |x|$ 的临界点,下列哪种说法正确?
1.1
$f$ has no critical point because $f'(0)$ is undefined$f$ 没有临界点,因为 $f'(0)$ 无定义
$x=0$ is a critical point because $f'(0) = 0$$x=0$ 是临界点,因为 $f'(0) = 0$
$f$ has a critical point at every integer$f$ 在每个整数处都有临界点
$x=0$ is a critical point because $f'(0)$ does not exist$x=0$ 是临界点,因为 $f'(0)$ 不存在
Correct. $f'(x) = -1$ for $x<0$ and $+1$ for $x>0$, so $f'(0)$ fails to exist while $0$ is in the domain. That makes $0$ a critical point, and indeed the absolute minimum.正确。$x<0$ 时 $f'(x) = -1$,$x>0$ 时为 $+1$,故 $f'(0)$ 不存在,而 $0$ 在定义域内。于是 $0$ 是临界点,事实上正是绝对最小值点。
A critical point occurs where $f'$ is zero or undefined. Here $f'(0)$ does not exist, so the correct option is the one citing the undefined derivative.临界点出现在 $f'$ 为零或无定义之处。这里 $f'(0)$ 不存在,所以正确选项是指出导数不存在的那一个。

The Mean Value Theorem中值定理

Key idea. The Mean Value Theorem converts a global statement about average rate of change into a pointwise statement about an instantaneous rate. It is the engine behind the sign analysis that governs the rest of the unit: it is exactly why a derivative that is positive on an interval forces the function to increase there.
核心思想。中值定理(Mean Value Theorem / MVT)把关于平均变化率的整体性结论,转化为关于某一点瞬时变化率的逐点结论。它是本单元后续所有正负号分析的引擎:正是它解释了为什么导数在某区间上为正就迫使函数在该区间上递增。

Theorem (Rolle's Theorem). Suppose $f$ is continuous on $[a,b]$, differentiable on $(a,b)$, and $f(a) = f(b)$. Then there exists $c \in (a,b)$ with $f'(c) = 0$.

定理(罗尔定理,Rolle's Theorem)。设 $f$ 在 $[a,b]$ 上连续,在 $(a,b)$ 上可微,且 $f(a) = f(b)$。则存在 $c \in (a,b)$ 使 $f'(c) = 0$。

Theorem (Mean Value Theorem). Suppose $f$ is continuous on $[a,b]$ and differentiable on $(a,b)$. Then there exists $c \in (a,b)$ such that the instantaneous rate at $c$ equals the average rate over $[a,b]$.

定理(中值定理,Mean Value Theorem)。设 $f$ 在 $[a,b]$ 上连续,在 $(a,b)$ 上可微。则存在 $c \in (a,b)$,使 $c$ 处的瞬时变化率等于 $[a,b]$ 上的平均变化率。

Mean Value Theorem中值定理(Mean Value Theorem)
$$ f'(c) = \frac{f(b) - f(a)}{b - a}. $$

Corollary. If $f'(x) = 0$ for all $x$ in an interval $I$, then $f$ is constant on $I$. Consequently, if $f'(x) = g'(x)$ throughout $I$, then $f - g$ is constant on $I$. This is the uniqueness statement that makes antiderivatives well defined up to a constant.

推论。若在区间 $I$ 上处处 $f'(x) = 0$,则 $f$ 在 $I$ 上为常数。于是若在 $I$ 上处处 $f'(x) = g'(x)$,则 $f - g$ 在 $I$ 上为常数。正是这条唯一性结论,使原函数(antiderivative)在相差一个常数的意义下被良好定义。

Worked Example 2.1: verifying the conclusion例题 2.1:验证结论

Find every $c$ guaranteed by the Mean Value Theorem for $f(x) = x^{3}$ on $[0,2]$.

The average rate of change is

$$ \frac{f(2) - f(0)}{2 - 0} = \frac{8 - 0}{2} = 4. $$

Set $f'(c) = 3c^{2}$ equal to $4$: $3c^{2} = 4$, so $c^{2} = 4/3$ and $c = 2/\sqrt{3} \approx 1.155$. The negative root is rejected because it lies outside $(0,2)$.

对 $f(x) = x^{3}$ 在 $[0,2]$ 上,求出中值定理所保证的所有 $c$。

平均变化率为

$$ \frac{f(2) - f(0)}{2 - 0} = \frac{8 - 0}{2} = 4. $$

令 $f'(c) = 3c^{2}$ 等于 $4$:$3c^{2} = 4$,故 $c^{2} = 4/3$,$c = 2/\sqrt{3} \approx 1.155$。负根因落在 $(0,2)$ 之外而舍去。

Going deeper: deriving the Mean Value Theorem from Rolle's Theorem深入一步:由罗尔定理推出中值定理

Given $f$ continuous on $[a,b]$ and differentiable on $(a,b)$, define the auxiliary function that subtracts the secant line.

$$ g(x) = f(x) - \left[ f(a) + \frac{f(b)-f(a)}{b-a}(x-a) \right]. $$

Then $g$ is continuous on $[a,b]$ and differentiable on $(a,b)$ as a difference of such functions. Direct substitution gives $g(a) = 0$ and $g(b) = 0$. By Rolle's Theorem there is a $c \in (a,b)$ with $g'(c) = 0$. Differentiating,

$$ g'(x) = f'(x) - \frac{f(b)-f(a)}{b-a}, \qquad g'(c) = 0 \implies f'(c) = \frac{f(b)-f(a)}{b-a}. $$

This is precisely the Mean Value Theorem, so the general theorem reduces to the special case $f(a)=f(b)$.

设 $f$ 在 $[a,b]$ 上连续、在 $(a,b)$ 上可微,定义减去割线的辅助函数。

$$ g(x) = f(x) - \left[ f(a) + \frac{f(b)-f(a)}{b-a}(x-a) \right]. $$

作为这类函数之差,$g$ 在 $[a,b]$ 上连续、在 $(a,b)$ 上可微。直接代入得 $g(a) = 0$ 与 $g(b) = 0$。由罗尔定理,存在 $c \in (a,b)$ 使 $g'(c) = 0$。求导得

$$ g'(x) = f'(x) - \frac{f(b)-f(a)}{b-a}, \qquad g'(c) = 0 \implies f'(c) = \frac{f(b)-f(a)}{b-a}. $$

这正是中值定理,于是一般情形归结为特例 $f(a)=f(b)$。

Worked Example 2.2: a bound from the Mean Value Theorem例题 2.2:用中值定理得到估界

Suppose $f$ is differentiable on $\mathbb{R}$ with $f(0) = 3$ and $|f'(x)| \le 2$ for every $x$. How large can $f(5)$ possibly be?

Apply the Mean Value Theorem on $[0,5]$: there is a $c \in (0,5)$ with

$$ f(5) - f(0) = f'(c)\,(5 - 0) = 5\,f'(c). $$

Therefore $|f(5) - 3| = 5\,|f'(c)| \le 5 \cdot 2 = 10$, which gives $-7 \le f(5) \le 13$. The largest possible value is $f(5) = 13$, attained when $f'(c) = 2$ throughout, that is when $f(x) = 2x + 3$. This is the model argument for turning a derivative bound into a bound on the function itself, and it underlies error estimates throughout calculus.

设 $f$ 在 $\mathbb{R}$ 上可微,$f(0) = 3$,且对每个 $x$ 有 $|f'(x)| \le 2$。$f(5)$ 最大可能是多少?

在 $[0,5]$ 上用中值定理:存在 $c \in (0,5)$ 使

$$ f(5) - f(0) = f'(c)\,(5 - 0) = 5\,f'(c). $$

因此 $|f(5) - 3| = 5\,|f'(c)| \le 5 \cdot 2 = 10$,即 $-7 \le f(5) \le 13$。最大可能值为 $f(5) = 13$,当 $f'(c)$ 全程等于 $2$(即 $f(x) = 2x + 3$)时取得。这是把导数估界转化为函数本身估界的范本论证,也是整个微积分中误差估计的基础。

Worked Example 2.3: proving an inequality with the Mean Value Theorem例题 2.3:用中值定理证明不等式

Prove that $\sin b - \sin a \le b - a$ for all $a \le b$.

If $a = b$ both sides are zero. For $a < b$, apply the Mean Value Theorem to $f(x) = \sin x$ on $[a,b]$: there is a $c \in (a,b)$ with

$$ \frac{\sin b - \sin a}{b - a} = \cos c. $$

Since $\cos c \le 1$ and $b - a > 0$, multiplying gives $\sin b - \sin a \le b - a$. The same argument with $|\cos c| \le 1$ yields the sharper statement $|\sin b - \sin a| \le |b - a|$, which says that sine is a nonexpansive (Lipschitz) function with constant $1$.

证明:对一切 $a \le b$ 有 $\sin b - \sin a \le b - a$。

若 $a = b$,两边均为零。若 $a < b$,对 $f(x) = \sin x$ 在 $[a,b]$ 上用中值定理:存在 $c \in (a,b)$ 使

$$ \frac{\sin b - \sin a}{b - a} = \cos c. $$

由 $\cos c \le 1$ 且 $b - a > 0$,两边相乘得 $\sin b - \sin a \le b - a$。用 $|\cos c| \le 1$ 作同样论证可得更强的 $|\sin b - \sin a| \le |b - a|$,即正弦是常数为 $1$ 的非扩张(利普希茨,Lipschitz)函数。

Common error. The Mean Value Theorem is an existence statement: it guarantees that at least one such $c$ exists, but it never tells you which one without further work, and there may be several. Students often misread the conclusion as "$f'(c)$ equals the average for every $c$", or they apply the theorem when a hypothesis fails. Continuity on the closed $[a,b]$ and differentiability on the open $(a,b)$ are both required; dropping either one breaks the guarantee, as $f(x)=|x|$ on $[-1,1]$ shows. Verify both hypotheses before invoking the conclusion.
常见错误。中值定理是存在性命题:它只保证至少存在一个这样的 $c$,不经进一步计算并不告诉你是哪一个,而且可能有好几个。学生常把结论误读成"对每个 $c$ 都有 $f'(c)$ 等于平均率",或在前提不满足时硬套定理。闭区间 $[a,b]$ 上的连续性开区间 $(a,b)$ 上的可微性二者都必须满足;缺一就破坏保证,如 $f(x)=|x|$ 在 $[-1,1]$ 上所示。引用结论前请先核验两个前提。
The Mean Value Theorem fails to apply to $f(x) = |x|$ on $[-1,1]$ because中值定理不能用于 $f(x) = |x|$ 在 $[-1,1]$ 上,原因是
2.1
$f$ is not continuous on $[-1,1]$$f$ 在 $[-1,1]$ 上不连续
$f(-1) \ne f(1)$$f(-1) \ne f(1)$
$f$ is not differentiable at $0 \in (-1,1)$$f$ 在 $0 \in (-1,1)$ 处不可微
the interval is not closed该区间不是闭区间
Correct. $f$ is continuous on $[-1,1]$, but differentiability on the open interval is required and fails at $0$. The hypothesis is violated, so the theorem gives no guarantee.正确。$f$ 在 $[-1,1]$ 上连续,但定理要求在开区间上可微,而它在 $0$ 处不可微。前提被破坏,定理因此不给出任何保证。
$f(x)=|x|$ is continuous everywhere and $f(-1)=f(1)=1$. The breakdown is the missing derivative at the corner $x=0$.$f(x)=|x|$ 处处连续,且 $f(-1)=f(1)=1$。问题出在尖角 $x=0$ 处导数不存在。

Increasing, Decreasing, and the First Derivative Test单调递增、递减与一阶导数判别法

Key idea. The sign of $f'$ records the direction of motion of the graph. By examining how that sign changes as $x$ passes through each critical point, the First Derivative Test classifies every critical point as a local maximum, a local minimum, or neither, using only first-order information.
核心思想。$f'$ 的正负号记录了图像的运动方向。通过考察当 $x$ 经过每个临界点时该符号如何变化,一阶导数判别法(First Derivative Test)仅凭一阶信息就能把每个临界点判为局部最大值、局部最小值或两者皆非。

Theorem (Increasing/Decreasing Test). Let $f$ be continuous on $[a,b]$ and differentiable on $(a,b)$. If $f'(x) > 0$ for all $x$ in $(a,b)$, then $f$ is increasing on $[a,b]$. If $f'(x) < 0$ throughout, then $f$ is decreasing on $[a,b]$. Each direction follows directly from the Mean Value Theorem applied to any two points of the interval.

定理(单调性判别法,Increasing/Decreasing Test)。设 $f$ 在 $[a,b]$ 上连续、在 $(a,b)$ 上可微。若对 $(a,b)$ 内一切 $x$ 有 $f'(x) > 0$,则 $f$ 在 $[a,b]$ 上递增;若处处 $f'(x) < 0$,则在 $[a,b]$ 上递减。两个方向都可由对区间内任意两点应用中值定理直接得到。

First Derivative Test at a critical point $c$临界点 $c$ 处的一阶导数判别法(First Derivative Test)
$$ \begin{aligned} f' : (+) \to (-) \text{ across } c &\implies \text{local maximum at } c, \\ f' : (-) \to (+) \text{ across } c &\implies \text{local minimum at } c, \\ f' \text{ no sign change} &\implies \text{neither.} \end{aligned} $$

Procedure. List all critical points, mark them on a number line, and test the sign of $f'$ on each resulting open interval by evaluating $f'$ at a convenient point. The sign pattern then determines both the monotonicity intervals and the classification of each critical point.

步骤。列出所有临界点,在数轴上标出它们,然后在每个由此分出的开区间内取一个方便的点代入 $f'$ 以判断符号。所得的符号分布既决定单调区间,也决定每个临界点的分类。

Worked Example 3.1: a full sign analysis例题 3.1:完整的符号分析

Classify the critical points of $f(x) = x^{3} - 3x^{2} - 9x + 5$ from Example 1.1, where $f'(x) = 3(x-3)(x+1)$.

The critical points $x = -1$ and $x = 3$ split the line into three intervals. Test one point in each.

$$ f'(-2) = 3(-5)(-1) > 0, \quad f'(0) = 3(-3)(1) < 0, \quad f'(4) = 3(1)(5) > 0. $$

So $f$ increases on $(-\infty,-1)$, decreases on $(-1,3)$, and increases on $(3,\infty)$. The sign goes $(+)\to(-)$ at $x=-1$, giving a local maximum, and $(-)\to(+)$ at $x=3$, giving a local minimum.

对例题 1.1 中的 $f(x) = x^{3} - 3x^{2} - 9x + 5$(其中 $f'(x) = 3(x-3)(x+1)$)的临界点进行分类。

临界点 $x = -1$ 与 $x = 3$ 把数轴分成三个区间。每个区间各取一点检验。

$$ f'(-2) = 3(-5)(-1) > 0, \quad f'(0) = 3(-3)(1) < 0, \quad f'(4) = 3(1)(5) > 0. $$

于是 $f$ 在 $(-\infty,-1)$ 上递增,在 $(-1,3)$ 上递减,在 $(3,\infty)$ 上递增。$x=-1$ 处符号 $(+)\to(-)$,给出局部最大值;$x=3$ 处 $(-)\to(+)$,给出局部最小值。

Worked Example 3.2: a critical point where $f'$ is undefined例题 3.2:$f'$ 不存在处的临界点

Classify the critical points of $f(x) = x^{2/3}(x - 5)$ and find the intervals of increase and decrease.

Write $f(x) = x^{5/3} - 5x^{2/3}$ and differentiate.

$$ f'(x) = \tfrac{5}{3}x^{2/3} - \tfrac{10}{3}x^{-1/3} = \frac{5}{3}\,\frac{x - 2}{x^{1/3}}. $$

The derivative is zero at $x = 2$ and undefined at $x = 0$ (which is in the domain), so the critical points are $x = 0$ and $x = 2$. Test the sign of $f'$ on the three intervals, watching the sign of both the numerator $x-2$ and the denominator $x^{1/3}$:

$$ f'(-1) = \tfrac{5}{3}\cdot\tfrac{-3}{-1} > 0, \quad f'(1) = \tfrac{5}{3}\cdot\tfrac{-1}{1} < 0, \quad f'(8) = \tfrac{5}{3}\cdot\tfrac{6}{2} > 0. $$

So $f$ increases on $(-\infty,0)$, decreases on $(0,2)$, and increases on $(2,\infty)$. The sign change $(+)\to(-)$ at $x=0$ gives a local maximum (a cusp, since $f'$ is undefined there), and $(-)\to(+)$ at $x=2$ gives a local minimum. This shows why the sign chart must include points where $f'$ is undefined, not only its roots.

对 $f(x) = x^{2/3}(x - 5)$ 的临界点分类,并求出递增、递减区间。

写成 $f(x) = x^{5/3} - 5x^{2/3}$ 再求导。

$$ f'(x) = \tfrac{5}{3}x^{2/3} - \tfrac{10}{3}x^{-1/3} = \frac{5}{3}\,\frac{x - 2}{x^{1/3}}. $$

导数在 $x = 2$ 处为零,在 $x = 0$ 处不存在(而 $0$ 在定义域内),故临界点为 $x = 0$ 与 $x = 2$。在三个区间上检验 $f'$ 的符号,同时关注分子 $x-2$ 与分母 $x^{1/3}$ 的符号:

$$ f'(-1) = \tfrac{5}{3}\cdot\tfrac{-3}{-1} > 0, \quad f'(1) = \tfrac{5}{3}\cdot\tfrac{-1}{1} < 0, \quad f'(8) = \tfrac{5}{3}\cdot\tfrac{6}{2} > 0. $$

故 $f$ 在 $(-\infty,0)$ 上递增,在 $(0,2)$ 上递减,在 $(2,\infty)$ 上递增。$x=0$ 处符号变化 $(+)\to(-)$ 给出局部最大值(一个尖点,因为 $f'$ 在此不存在),$x=2$ 处 $(-)\to(+)$ 给出局部最小值。这说明符号表必须包含 $f'$ 不存在的点,而不只是它的根。

Worked Example 3.3: increase and decrease for a product with an exponential例题 3.3:含指数因子之积的单调性

Find where $f(x) = x e^{-x}$ is increasing or decreasing, and classify any local extremum.

By the product rule,

$$ f'(x) = e^{-x} - x e^{-x} = (1 - x)e^{-x}. $$

Because $e^{-x} > 0$ for every $x$, the sign of $f'$ matches the sign of $1 - x$. Hence $f' > 0$ on $(-\infty,1)$ and $f' < 0$ on $(1,\infty)$: the function increases then decreases. The sign change $(+)\to(-)$ at $x = 1$ gives a local maximum, with value $f(1) = e^{-1} \approx 0.368$. Since this is the only critical point and the function rises to it then falls forever after, it is in fact the absolute maximum on the whole line.

求 $f(x) = x e^{-x}$ 的递增、递减区间,并对任何局部极值进行分类。

由乘积法则(Product Rule),

$$ f'(x) = e^{-x} - x e^{-x} = (1 - x)e^{-x}. $$

由于对每个 $x$ 都有 $e^{-x} > 0$,$f'$ 的符号与 $1 - x$ 一致。故 $f'$ 在 $(-\infty,1)$ 上为正、在 $(1,\infty)$ 上为负:函数先增后减。$x = 1$ 处符号变化 $(+)\to(-)$ 给出局部最大值,其值 $f(1) = e^{-1} \approx 0.368$。由于这是唯一的临界点,且函数先升到它、此后永远下降,它实际上是整条数轴上的绝对最大值。

Common error. The First Derivative Test classifies a critical point by the change of sign of $f'$, not by the value of $f'$ at the point (which is zero or undefined). A repeated even factor such as $(x-2)^2$ does not change sign, so the point is neither a max nor a min even though $f'=0$ there. A related mistake is to forget the points where $f'$ is undefined when building the sign chart: those split the domain just as roots of $f'$ do, and skipping them can miss a genuine extremum like the cusp in Example 3.2.
常见错误。一阶导数判别法是依据 $f'$ 的符号变化来分类临界点,而非依据 $f'$ 在该点的取值(它本就是零或不存在)。像 $(x-2)^2$ 这样的偶次重因子不改变符号,所以即便 $f'=0$,该点也既非最大也非最小。相关的错误是建符号表时忘掉 $f'$ 不存在的点:它们和 $f'$ 的根一样把定义域分段,漏掉它们就可能错过例题 3.2 那样真正的极值(尖点)。
If $f'(x) = (x-2)^{2}(x+1)$, what happens at $x = 2$?若 $f'(x) = (x-2)^{2}(x+1)$,在 $x = 2$ 处会怎样?
3.1
Local minimum, since $f'(2) = 0$局部最小值,因为 $f'(2) = 0$
Neither, because $f'$ does not change sign at $x=2$两者皆非,因为 $f'$ 在 $x=2$ 处不改变符号
Local maximum, since the factor is squared局部最大值,因为该因子是平方
It cannot be a critical point它不可能是临界点
Correct. The factor $(x-2)^{2}$ is nonnegative, so near $x=2$ the sign of $f'$ is governed by $(x+1) > 0$ on both sides. With no sign change, the First Derivative Test gives neither a max nor a min.正确。因子 $(x-2)^{2}$ 非负,故在 $x=2$ 附近 $f'$ 的符号由两侧均为正的 $(x+1)$ 决定。没有符号变化,一阶导数判别法给出既非最大也非最小。
A repeated even factor does not change sign across its root. Since $f'$ stays positive on both sides of $x=2$, the point is neither a local max nor a local min.偶次重因子在其根两侧不改变符号。由于 $f'$ 在 $x=2$ 两侧都为正,该点既非局部最大也非局部最小。

Concavity and the Second Derivative Test凹凸性与二阶导数判别法

Key idea. The second derivative measures how the slope itself is changing, which is the bending of the graph. Where $f'' > 0$ the curve bends upward (concave up); where $f'' < 0$ it bends downward (concave down). Points where the concavity reverses are inflection points, and the second derivative gives a fast test for many critical points.
核心思想。二阶导数衡量斜率本身的变化,也就是图像的弯曲。$f'' > 0$ 处曲线向上弯(凹向上,concave up);$f'' < 0$ 处向下弯(凹向下,concave down)。凹凸性反转之处就是拐点(inflection point),而二阶导数为许多临界点提供了一种快速判别法。

Definition. $f$ is concave up on an interval if $f'$ is increasing there, and concave down if $f'$ is decreasing there. A point on the graph where the concavity changes is an inflection point.

定义。若 $f'$ 在某区间上递增,则称 $f$ 在该区间上凹向上concave up);若 $f'$ 在该区间上递减,则凹向下concave down)。图像上凹凸性发生改变的点称为拐点inflection point)。

Theorem (Concavity Test). If $f''(x) > 0$ for all $x$ in an interval, then $f$ is concave up there; if $f''(x) < 0$ throughout, then $f$ is concave down there.

定理(凹凸性判别法,Concavity Test)。若在某区间内对一切 $x$ 有 $f''(x) > 0$,则 $f$ 在该区间凹向上;若处处 $f''(x) < 0$,则凹向下。

Second Derivative Test at a critical point $c$ with $f'(c)=0$满足 $f'(c)=0$ 的临界点 $c$ 处的二阶导数判别法(Second Derivative Test)
$$ f''(c) > 0 \implies \text{local minimum}, \qquad f''(c) < 0 \implies \text{local maximum}. $$

Remark. If $f''(c) = 0$ the Second Derivative Test is inconclusive and gives no information. The functions $x^{4}$ (minimum), $-x^{4}$ (maximum), and $x^{3}$ (neither) all have $f' = f'' = 0$ at the origin, so one must fall back on the First Derivative Test. A candidate inflection point likewise requires $f''$ to actually change sign, not merely to vanish.

注。若 $f''(c) = 0$,二阶导数判别法失效,不给出任何信息。函数 $x^{4}$(最小)、$-x^{4}$(最大)、$x^{3}$(皆非)在原点都有 $f' = f'' = 0$,所以只能退回到一阶导数判别法。同理,候选拐点也要求 $f''$ 真正改变符号,而不只是取零。

Worked Example 4.1: concavity and inflection例题 4.1:凹凸性与拐点

For $f(x) = x^{3} - 3x^{2} - 9x + 5$, analyze concavity. From $f'(x) = 3x^{2}-6x-9$,

$$ f''(x) = 6x - 6 = 6(x-1). $$

Thus $f'' < 0$ for $x < 1$ (concave down) and $f'' > 0$ for $x > 1$ (concave up). The concavity reverses at $x = 1$, where $f(1) = 1 - 3 - 9 + 5 = -6$, so $(1,-6)$ is an inflection point.

Checking the earlier critical points with the Second Derivative Test: $f''(-1) = -12 < 0$ confirms the local maximum, and $f''(3) = 12 > 0$ confirms the local minimum.

对 $f(x) = x^{3} - 3x^{2} - 9x + 5$ 分析凹凸性。由 $f'(x) = 3x^{2}-6x-9$,

$$ f''(x) = 6x - 6 = 6(x-1). $$

故 $x < 1$ 时 $f'' < 0$(凹向下),$x > 1$ 时 $f'' > 0$(凹向上)。凹凸性在 $x = 1$ 处反转,此处 $f(1) = 1 - 3 - 9 + 5 = -6$,所以 $(1,-6)$ 是拐点。

用二阶导数判别法核验先前的临界点:$f''(-1) = -12 < 0$ 证实局部最大值,$f''(3) = 12 > 0$ 证实局部最小值。

Going deeper: why $f'' > 0$ means concave up深入一步:为何 $f'' > 0$ 意味着凹向上

Suppose $f'' > 0$ on an interval $I$. By the Increasing/Decreasing Test applied to the function $f'$, the fact that $(f')' = f'' > 0$ forces $f'$ to be increasing on $I$. By definition, an increasing first derivative is exactly what concave up means: as $x$ grows, the tangent slope grows, so the curve turns upward and lies above each of its tangent lines on $I$. The concave down case is identical with inequalities reversed.

设在区间 $I$ 上 $f'' > 0$。对函数 $f'$ 应用单调性判别法,$(f')' = f'' > 0$ 迫使 $f'$ 在 $I$ 上递增。按定义,一阶导数递增正是凹向上的含义:随着 $x$ 增大,切线斜率增大,于是曲线向上弯,并在 $I$ 上位于其每条切线之上。凹向下的情形完全相同,只需把不等号反向。

Worked Example 4.2: a vanishing second derivative that is not an inflection例题 4.2:二阶导数为零却不是拐点

Examine $f(x) = x^4$ at the origin. Compute $f'(x) = 4x^3$ and $f''(x) = 12x^2$. Then $f''(0) = 0$, so one might suspect an inflection point. But $f''(x) = 12x^2 \ge 0$ for all $x$, so the concavity is up on both sides of $0$ and never reverses. There is no inflection point; the origin is in fact the absolute minimum. The lesson is that $f''(c) = 0$ is necessary but not sufficient for an inflection: the second derivative must actually change sign as $x$ crosses $c$.

考察 $f(x) = x^4$ 在原点。算得 $f'(x) = 4x^3$,$f''(x) = 12x^2$。则 $f''(0) = 0$,乍看像是拐点。但对一切 $x$ 都有 $f''(x) = 12x^2 \ge 0$,所以 $0$ 两侧都凹向上,从不反转。这里没有拐点;原点其实是绝对最小值。要点是 $f''(c) = 0$ 对拐点是必要而非充分条件:当 $x$ 跨过 $c$ 时,二阶导数必须真正改变符号。

Worked Example 4.3: the Second Derivative Test on a transcendental function例题 4.3:超越函数上的二阶导数判别法

Classify the critical points of $f(x) = 2\cos x + \cos 2x$ on $(0, 2\pi)$ using the Second Derivative Test, and locate the inflection points.

Differentiate twice.

$$ f'(x) = -2\sin x - 2\sin 2x = -2\sin x(1 + 2\cos x), \qquad f''(x) = -2\cos x - 4\cos 2x. $$

Setting $f'(x) = 0$ gives $\sin x = 0$ or $\cos x = -\tfrac12$. On $(0,2\pi)$ this yields $x = \pi$ (from $\sin x = 0$) and $x = \tfrac{2\pi}{3},\ \tfrac{4\pi}{3}$ (from $\cos x = -\tfrac12$). Evaluate $f''$:

$$ f''\!\left(\tfrac{2\pi}{3}\right) = -2(-\tfrac12) - 4(-\tfrac12) = 1 + 2 = 3 > 0 \ \text{(local min)}, $$ $$ f''(\pi) = -2(-1) - 4(1) = 2 - 4 = -2 < 0 \ \text{(local max)}, $$ $$ f''\!\left(\tfrac{4\pi}{3}\right) = 3 > 0 \ \text{(local min, by symmetry)}. $$

So the test cleanly classifies all three critical points, illustrating how much faster the Second Derivative Test can be than a full sign chart when $f''$ is easy to evaluate.

用二阶导数判别法对 $f(x) = 2\cos x + \cos 2x$ 在 $(0, 2\pi)$ 上的临界点分类,并找出拐点。

求两次导。

$$ f'(x) = -2\sin x - 2\sin 2x = -2\sin x(1 + 2\cos x), \qquad f''(x) = -2\cos x - 4\cos 2x. $$

令 $f'(x) = 0$ 得 $\sin x = 0$ 或 $\cos x = -\tfrac12$。在 $(0,2\pi)$ 上由 $\sin x = 0$ 得 $x = \pi$,由 $\cos x = -\tfrac12$ 得 $x = \tfrac{2\pi}{3},\ \tfrac{4\pi}{3}$。代入 $f''$:

$$ f''\!\left(\tfrac{2\pi}{3}\right) = -2(-\tfrac12) - 4(-\tfrac12) = 1 + 2 = 3 > 0 \ \text{(local min)}, $$ $$ f''(\pi) = -2(-1) - 4(1) = 2 - 4 = -2 < 0 \ \text{(local max)}, $$ $$ f''\!\left(\tfrac{4\pi}{3}\right) = 3 > 0 \ \text{(local min, by symmetry)}. $$

于是该判别法干净利落地分类了全部三个临界点,说明当 $f''$ 容易求值时,二阶导数判别法比画完整符号表快得多。

Common error. A point with $f''(c) = 0$ is only a candidate inflection point; you must verify that $f''$ changes sign there. Treating every root of $f''$ as an inflection point is wrong, as $x^4$ shows. The opposite error also appears: an inflection point can occur where $f''$ is undefined (for example $f(x)=x^{1/3}$ at the origin), so checking only the zeros of $f''$ can miss one. Build the concavity sign chart from both the zeros and the undefined points of $f''$, exactly as you do for $f'$.
常见错误。满足 $f''(c) = 0$ 的点只是候选拐点;你必须核验 $f''$ 在此确实变号。把 $f''$ 的每个根都当成拐点是错的,如 $x^4$ 所示。相反的错误也会出现:拐点也可能出现在 $f''$ 不存在之处(例如 $f(x)=x^{1/3}$ 在原点),所以只查 $f''$ 的零点会漏掉它。要像对 $f'$ 那样,用 $f''$ 的零点与不存在点共同建立凹凸性符号表。
At a critical point $c$ where $f'(c)=0$ and $f''(c)=0$, the Second Derivative Test在满足 $f'(c)=0$ 且 $f''(c)=0$ 的临界点 $c$ 处,二阶导数判别法
4.1
is inconclusive and another method is needed无法判定,需要另用其他方法
guarantees an inflection point at $c$保证 $c$ 处是拐点
guarantees a local maximum保证是局部最大值
guarantees a local minimum保证是局部最小值
Correct. When $f''(c)=0$ the test gives no conclusion; $x^4$, $-x^4$, and $x^3$ all satisfy $f'(0)=f''(0)=0$ yet behave differently. Fall back on the First Derivative Test.正确。当 $f''(c)=0$ 时该判别法无结论;$x^4$、$-x^4$、$x^3$ 都满足 $f'(0)=f''(0)=0$ 却表现各异。应退回到一阶导数判别法。
A vanishing second derivative does not by itself signal a max, a min, or even an inflection point. The test is simply inconclusive.二阶导数为零本身既不标志最大、最小,也不标志拐点。该判别法就是无法判定。

Curve Sketching曲线描绘

Key idea. A faithful sketch is assembled by collecting independent pieces of information, domain, symmetry, intercepts, asymptotes, monotonicity, and concavity, and overlaying them on one set of axes. No single derivative tells the whole story; the shape emerges from their intersection.
核心思想。一张忠实的草图是把若干彼此独立的信息——定义域、对称性、截距、渐近线(asymptote)、单调性、凹凸性——收集起来,叠加在同一坐标系上拼出来的。没有哪一个导数能道尽全貌;形状是在它们的交汇处浮现的。

Procedure (a checklist for sketching $y = f(x)$).

1. Determine the domain and any symmetry (even, odd, periodic). 2. Find the $x$- and $y$-intercepts when feasible. 3. Find vertical asymptotes (where $f \to \pm\infty$) and horizontal asymptotes from the limits at infinity. 4. Use $f'$ for intervals of increase and decrease and for local extrema. 5. Use $f''$ for concavity and inflection points. 6. Plot the key points and connect them respecting the established slope and bending.

步骤(描绘 $y = f(x)$ 的清单)。

1. 确定定义域与对称性(偶、奇、周期)。2. 在可行时求 $x$ 截距与 $y$ 截距。3. 求竖直渐近线($f \to \pm\infty$ 处)以及由无穷远处极限(limit)得到的水平渐近线(horizontal asymptote)。4. 用 $f'$ 求递增、递减区间与局部极值。5. 用 $f''$ 求凹凸性与拐点。6. 描出关键点,并按已确定的斜率与弯曲方向把它们连接起来。

Horizontal asymptote from a limit at infinity由无穷远处极限得到的水平渐近线(horizontal asymptote)
$$ \lim_{x \to \infty} f(x) = L \quad \text{or} \quad \lim_{x \to -\infty} f(x) = L \implies y = L \text{ is a horizontal asymptote.} $$
Worked Example 5.1: sketching a rational function例题 5.1:描绘有理函数

Sketch $f(x) = \dfrac{x^{2}}{x^{2} - 1}$.

Domain and symmetry. Defined for $x \ne \pm 1$. Since $f(-x) = f(x)$, the graph is even and symmetric about the $y$-axis.

Intercepts. $f(0) = 0$, so the only intercept is the origin.

Asymptotes. The denominator vanishes at $x = \pm 1$, giving vertical asymptotes there. For the end behavior,

$$ \lim_{x \to \pm\infty} \frac{x^{2}}{x^{2}-1} = 1, $$

so $y = 1$ is a horizontal asymptote.

Monotonicity. By the quotient rule,

$$ f'(x) = \frac{2x(x^{2}-1) - x^{2}(2x)}{(x^{2}-1)^{2}} = \frac{-2x}{(x^{2}-1)^{2}}. $$

The denominator is positive, so $f' > 0$ for $x < 0$ and $f' < 0$ for $x > 0$ (excluding the asymptotes). Thus $f$ increases on $(-\infty,-1)$ and $(-1,0)$ and decreases on $(0,1)$ and $(1,\infty)$, with a local maximum at the origin where $f(0)=0$. Combining these facts produces the characteristic three-branch graph hugging $y = 1$.

描绘 $f(x) = \dfrac{x^{2}}{x^{2} - 1}$。

定义域与对称性。在 $x \ne \pm 1$ 时有定义。由 $f(-x) = f(x)$ 知图像为偶函数,关于 $y$ 轴对称。

截距。$f(0) = 0$,故唯一的截距是原点。

渐近线。分母在 $x = \pm 1$ 处为零,给出该处的竖直渐近线。对于端点行为,

$$ \lim_{x \to \pm\infty} \frac{x^{2}}{x^{2}-1} = 1, $$

所以 $y = 1$ 是水平渐近线。

单调性。由商的法则(Quotient Rule),

$$ f'(x) = \frac{2x(x^{2}-1) - x^{2}(2x)}{(x^{2}-1)^{2}} = \frac{-2x}{(x^{2}-1)^{2}}. $$

分母为正,故 $x < 0$ 时 $f' > 0$,$x > 0$ 时 $f' < 0$(不含渐近线处)。于是 $f$ 在 $(-\infty,-1)$ 与 $(-1,0)$ 上递增,在 $(0,1)$ 与 $(1,\infty)$ 上递减,在原点处 $f(0)=0$ 取局部最大值。综合这些事实即得贴着 $y = 1$ 的特征性三支图像。

Worked Example 5.2: a slant asymptote例题 5.2:斜渐近线

Sketch $f(x) = \dfrac{x^{2} + 1}{x}$ and identify all asymptotes.

Domain and symmetry. Defined for $x \ne 0$. Since $f(-x) = -f(x)$, the graph is odd and symmetric about the origin.

Asymptotes. Long division gives $f(x) = x + \dfrac{1}{x}$. As $x \to 0$ the term $1/x$ dominates, so $x = 0$ is a vertical asymptote. As $x \to \pm\infty$ the term $1/x \to 0$, so the graph approaches the line $y = x$. This line is a slant (oblique) asymptote, which appears whenever the degree of the numerator is exactly one more than that of the denominator.

Monotonicity and shape. From $f'(x) = 1 - \dfrac{1}{x^{2}} = \dfrac{x^{2}-1}{x^{2}}$, the critical points are $x = \pm 1$. The derivative is positive for $|x| > 1$ and negative for $0 < |x| < 1$, so $x = -1$ is a local maximum with $f(-1) = -2$ and $x = 1$ is a local minimum with $f(1) = 2$. Notice the local minimum value $2$ exceeds the local maximum value $-2$; this is possible only because a vertical asymptote separates the two branches.

描绘 $f(x) = \dfrac{x^{2} + 1}{x}$ 并找出所有渐近线。

定义域与对称性。在 $x \ne 0$ 时有定义。由 $f(-x) = -f(x)$ 知图像为奇函数,关于原点对称。

渐近线。多项式除法给出 $f(x) = x + \dfrac{1}{x}$。当 $x \to 0$ 时 $1/x$ 项占主导,所以 $x = 0$ 是竖直渐近线。当 $x \to \pm\infty$ 时 $1/x \to 0$,所以图像趋近直线 $y = x$。这条直线是斜渐近线slant asymptote),只要分子次数恰好比分母次数高一次就会出现。

单调性与形状。由 $f'(x) = 1 - \dfrac{1}{x^{2}} = \dfrac{x^{2}-1}{x^{2}}$,临界点为 $x = \pm 1$。导数在 $|x| > 1$ 时为正、在 $0 < |x| < 1$ 时为负,故 $x = -1$ 是局部最大值 $f(-1) = -2$,$x = 1$ 是局部最小值 $f(1) = 2$。注意局部最小值 $2$ 竟大于局部最大值 $-2$;这只有在竖直渐近线把两支分开时才可能发生。

Worked Example 5.3: a full sketch combining $f'$ and $f''$例题 5.3:综合 $f'$ 与 $f''$ 的完整草图

Assemble the graph of $f(x) = x^{3} - 3x^{2} - 9x + 5$, the function studied throughout this unit.

Collecting the earlier results: $f$ increases on $(-\infty,-1)$, decreases on $(-1,3)$, and increases on $(3,\infty)$, with a local maximum at $(-1, 10)$ and a local minimum at $(3, -22)$. The concavity is down on $(-\infty,1)$ and up on $(1,\infty)$, with an inflection point at $(1,-6)$. The end behavior is $f \to -\infty$ as $x \to -\infty$ and $f \to +\infty$ as $x \to +\infty$, with no asymptotes since $f$ is a polynomial.

Plotting these features in order, rising to the crest at $(-1,10)$ while concave down, bending through the inflection at $(1,-6)$, sinking to the trough at $(3,-22)$ while concave up, then rising again, reproduces the standard cubic with one hump and one valley. The independent pieces, monotonicity from $f'$ and bending from $f''$, lock together into a single unambiguous shape.

拼出本单元始终研究的函数 $f(x) = x^{3} - 3x^{2} - 9x + 5$ 的图像。

汇总前面的结果:$f$ 在 $(-\infty,-1)$ 上递增,在 $(-1,3)$ 上递减,在 $(3,\infty)$ 上递增,在 $(-1, 10)$ 取局部最大值、在 $(3, -22)$ 取局部最小值。凹凸性在 $(-\infty,1)$ 上为凹向下、在 $(1,\infty)$ 上为凹向上,拐点在 $(1,-6)$。端点行为是 $x \to -\infty$ 时 $f \to -\infty$、$x \to +\infty$ 时 $f \to +\infty$,因 $f$ 是多项式而无渐近线。

按顺序画出这些特征:凹向下地升到 $(-1,10)$ 的峰顶,经 $(1,-6)$ 处拐点弯转,凹向上地降到 $(3,-22)$ 的谷底,再次上升,就重现了一峰一谷的标准三次曲线。两块独立的信息——来自 $f'$ 的单调性与来自 $f''$ 的弯曲——咬合成唯一确定的形状。

Common error. When the numerator and denominator of a rational function have the same degree, the horizontal asymptote is the ratio of leading coefficients, not $y = 0$. Reflexively writing $y = 0$ is the most common asymptote mistake. Two further pitfalls: a graph may cross its horizontal asymptote (the asymptote describes end behavior, not a barrier), and a slant asymptote appears, in place of a horizontal one, exactly when the numerator degree is one higher than the denominator degree. Always compare degrees before declaring the asymptote.
常见错误。当有理函数的分子与分母次数相同时,水平渐近线是首项系数之比,而非 $y = 0$。条件反射地写 $y = 0$ 是最常见的渐近线错误。还有两个陷阱:图像可以穿过它的水平渐近线(渐近线描述的是端点行为,不是屏障);当分子次数恰比分母高一次时,出现的是斜渐近线而非水平渐近线。下结论前务必先比较次数。
For $f(x) = \dfrac{3x^{2}+1}{x^{2}+4}$, the horizontal asymptote is对 $f(x) = \dfrac{3x^{2}+1}{x^{2}+4}$,水平渐近线是
5.1
$y = 0$$y = 0$
$y = 1/4$$y = 1/4$
$y = 3$$y = 3$
there is none不存在
Correct. The numerator and denominator have equal degree, so the limit at infinity is the ratio of leading coefficients, $3/1 = 3$. Hence $y = 3$ is the horizontal asymptote.正确。分子与分母次数相同,故无穷远处的极限是首项系数之比 $3/1 = 3$。因此 $y = 3$ 是水平渐近线。
When the top and bottom have the same degree, divide the leading coefficients. Here $3/1 = 3$, so the asymptote is $y=3$.分子分母次数相同时,取首项系数相除。这里 $3/1 = 3$,故渐近线为 $y=3$。

Optimization: The Closed Interval Method最优化:闭区间法

Key idea. On a closed bounded interval, the Extreme Value Theorem promises that the absolute extrema exist, and Fermat's Theorem tells us they live among the critical points and the two endpoints. The Closed Interval Method exploits this by simply evaluating $f$ at every candidate and reading off the largest and smallest values.
核心思想。在有界闭区间上,极值定理(Extreme Value Theorem)保证绝对极值存在,而费马定理告诉我们它们就藏在临界点与两个端点之中。闭区间法(Closed Interval Method)正是利用这一点:只需把 $f$ 在每个候选处求值,再读出最大与最小值即可。
Closed Interval Method for $f$ continuous on $[a,b]$$f$ 在 $[a,b]$ 上连续时的闭区间法(Closed Interval Method)
$$ \max_{[a,b]} f = \max\{\, f(\text{critical pts in }(a,b)),\, f(a),\, f(b) \,\}, $$ $$ \min_{[a,b]} f = \min\{\, f(\text{critical pts in }(a,b)),\, f(a),\, f(b) \,\}. $$

Procedure. 1. Confirm $f$ is continuous on $[a,b]$. 2. Find the critical points of $f$ in the open interval $(a,b)$. 3. Evaluate $f$ at those critical points and at both endpoints. 4. The largest of these values is the absolute maximum and the smallest is the absolute minimum. Sign analysis is unnecessary: only the list of function values matters.

步骤。1. 确认 $f$ 在 $[a,b]$ 上连续。2. 求 $f$ 在开区间 $(a,b)$ 内的临界点。3. 在这些临界点与两个端点处求 $f$ 的值。4. 其中最大的值是绝对最大值,最小的值是绝对最小值。无需做符号分析:只看这份函数值清单即可。

Worked Example 6.1: absolute extrema on a closed interval例题 6.1:闭区间上的绝对极值

Find the absolute maximum and minimum of $f(x) = x^{3} - 3x + 1$ on $[0, 3]$.

Differentiate: $f'(x) = 3x^{2} - 3 = 3(x-1)(x+1)$. The critical points are $x = \pm 1$, but only $x = 1$ lies in $(0,3)$, so the candidate list is $\{0, 1, 3\}$.

$$ f(0) = 1, \qquad f(1) = 1 - 3 + 1 = -1, \qquad f(3) = 27 - 9 + 1 = 19. $$

The absolute maximum is $19$, attained at $x = 3$ (an endpoint), and the absolute minimum is $-1$, attained at the interior critical point $x = 1$.

求 $f(x) = x^{3} - 3x + 1$ 在 $[0, 3]$ 上的绝对最大值与最小值。

求导:$f'(x) = 3x^{2} - 3 = 3(x-1)(x+1)$。临界点为 $x = \pm 1$,但只有 $x = 1$ 落在 $(0,3)$ 内,故候选名单为 $\{0, 1, 3\}$。

$$ f(0) = 1, \qquad f(1) = 1 - 3 + 1 = -1, \qquad f(3) = 27 - 9 + 1 = 19. $$

绝对最大值为 $19$,在端点 $x = 3$ 处取得;绝对最小值为 $-1$,在内部临界点 $x = 1$ 处取得。

Worked Example 6.2: a critical point with an undefined derivative例题 6.2:导数不存在的临界点

Find the absolute extrema of $f(x) = x^{2/3}(5 - x)$ on $[-1, 4]$.

Write $f(x) = 5x^{2/3} - x^{5/3}$, so $f'(x) = \tfrac{10}{3}x^{-1/3} - \tfrac{5}{3}x^{2/3} = \dfrac{5}{3}\cdot\dfrac{2 - x}{x^{1/3}}$, which is zero at $x = 2$ and undefined at $x = 0$. Both lie in $(-1,4)$, so the candidate list is $\{-1,\ 0,\ 2,\ 4\}$: two endpoints and two critical points.

$$ f(-1) = 1\cdot 6 = 6, \quad f(0) = 0, \quad f(2) = 2^{2/3}(3) \approx 4.76, \quad f(4) = 4^{2/3}(1) \approx 2.52. $$

The absolute maximum is $6$ at the endpoint $x = -1$, and the absolute minimum is $0$ at the interior critical point $x = 0$ where the derivative is undefined. Had we tested only the zeros of $f'$, the true minimum would have been missed.

求 $f(x) = x^{2/3}(5 - x)$ 在 $[-1, 4]$ 上的绝对极值。

写成 $f(x) = 5x^{2/3} - x^{5/3}$,则 $f'(x) = \tfrac{10}{3}x^{-1/3} - \tfrac{5}{3}x^{2/3} = \dfrac{5}{3}\cdot\dfrac{2 - x}{x^{1/3}}$,它在 $x = 2$ 处为零、在 $x = 0$ 处不存在。两者都落在 $(-1,4)$ 内,故候选名单为 $\{-1,\ 0,\ 2,\ 4\}$:两个端点加两个临界点。

$$ f(-1) = 1\cdot 6 = 6, \quad f(0) = 0, \quad f(2) = 2^{2/3}(3) \approx 4.76, \quad f(4) = 4^{2/3}(1) \approx 2.52. $$

绝对最大值为 $6$,在端点 $x = -1$ 处;绝对最小值为 $0$,在导数不存在的内部临界点 $x = 0$ 处。若只检验 $f'$ 的零点,就会漏掉真正的最小值。

Worked Example 6.3: a trigonometric extremum on a closed interval例题 6.3:闭区间上的三角函数极值

Find the absolute extrema of $f(x) = x - 2\sin x$ on $[0, 2\pi]$, continuing Example 1.3.

The interior critical points are $x = \tfrac{\pi}{3}$ and $x = \tfrac{5\pi}{3}$, so the candidate list is $\{0,\ \tfrac{\pi}{3},\ \tfrac{5\pi}{3},\ 2\pi\}$. Evaluate, using $\sin\tfrac{\pi}{3} = \tfrac{\sqrt3}{2}$ and $\sin\tfrac{5\pi}{3} = -\tfrac{\sqrt3}{2}$:

$$ f(0) = 0, \quad f\!\left(\tfrac{\pi}{3}\right) = \tfrac{\pi}{3} - \sqrt3 \approx -0.68, \quad f\!\left(\tfrac{5\pi}{3}\right) = \tfrac{5\pi}{3} + \sqrt3 \approx 6.97, \quad f(2\pi) = 2\pi \approx 6.28. $$

The absolute maximum is $\tfrac{5\pi}{3} + \sqrt3$ at the interior critical point $x = \tfrac{5\pi}{3}$, and the absolute minimum is $\tfrac{\pi}{3} - \sqrt3$ at $x = \tfrac{\pi}{3}$. Note the maximum is interior and exceeds both endpoint values, which is why endpoints alone never suffice.

承接例题 1.3,求 $f(x) = x - 2\sin x$ 在 $[0, 2\pi]$ 上的绝对极值。

内部临界点为 $x = \tfrac{\pi}{3}$ 与 $x = \tfrac{5\pi}{3}$,故候选名单为 $\{0,\ \tfrac{\pi}{3},\ \tfrac{5\pi}{3},\ 2\pi\}$。利用 $\sin\tfrac{\pi}{3} = \tfrac{\sqrt3}{2}$ 与 $\sin\tfrac{5\pi}{3} = -\tfrac{\sqrt3}{2}$ 求值:

$$ f(0) = 0, \quad f\!\left(\tfrac{\pi}{3}\right) = \tfrac{\pi}{3} - \sqrt3 \approx -0.68, \quad f\!\left(\tfrac{5\pi}{3}\right) = \tfrac{5\pi}{3} + \sqrt3 \approx 6.97, \quad f(2\pi) = 2\pi \approx 6.28. $$

绝对最大值为 $\tfrac{5\pi}{3} + \sqrt3$,在内部临界点 $x = \tfrac{5\pi}{3}$ 处;绝对最小值为 $\tfrac{\pi}{3} - \sqrt3$,在 $x = \tfrac{\pi}{3}$ 处。注意最大值在内部,且超过两个端点的值,这正是为何仅靠端点永远不够。

Common error. The Closed Interval Method asks for the largest and smallest function values, so report $f(c)$, not the location $c$. A second slip is to discard critical points that fall outside $(a,b)$ but then forget to include the endpoints, or to keep a critical point that the problem's domain excludes. Finally, the method requires $f$ to be continuous on $[a,b]$: a discontinuity inside the interval can produce a supremum that is never attained, and then comparing a finite candidate list is not valid.
常见错误。闭区间法求的是最大与最小的函数值,所以要报告 $f(c)$,而不是位置 $c$。第二个疏漏是舍去落在 $(a,b)$ 之外的临界点后却忘了把端点纳入,或者保留了被题目定义域排除的临界点。最后,该方法要求 $f$ 在 $[a,b]$ 上连续:区间内部的间断可能产生一个永远取不到的上确界,此时比较一份有限候选清单就不再成立。
In the Closed Interval Method on $[a,b]$, which set of points must be tested?在 $[a,b]$ 上的闭区间法中,必须检验哪一组点?
6.1
Only the points where $f'(x)=0$只检验 $f'(x)=0$ 的点
Only the endpoints $a$ and $b$只检验端点 $a$ 与 $b$
Only the points where $f''(x)=0$只检验 $f''(x)=0$ 的点
All critical points in $(a,b)$ together with the endpoints $a$ and $b$$(a,b)$ 内的所有临界点连同端点 $a$ 与 $b$
Correct. The absolute extrema occur either at an interior critical point or at an endpoint, so the full candidate list is the critical points in $(a,b)$ plus $a$ and $b$.正确。绝对极值要么出现在内部临界点,要么在端点,故完整候选名单是 $(a,b)$ 内的临界点加上 $a$ 与 $b$。
Endpoints can host the extreme values too, and critical points include those where $f'$ is undefined, not only where $f'=0$. Test all critical points in $(a,b)$ plus both endpoints.端点也可能承载极值,而临界点包括 $f'$ 不存在之处,不只是 $f'=0$ 之处。要检验 $(a,b)$ 内所有临界点加两个端点。

Applied Optimization应用最优化

Key idea. Applied optimization translates a word problem into a function of one variable on a feasible domain, then applies the calculus of extrema. The art is the modeling step: name the quantity to optimize, express it through a constraint in a single variable, and identify the domain dictated by the physical situation.
核心思想。应用最优化(optimization)是把一道应用题在某个可行定义域上化为单变量函数,再施以求极值的微积分。功夫在建模这一步:给要优化的量命名,借助约束把它表示成单一变量的函数,并辨认出由实际情形决定的定义域。

Procedure. 1. Identify the quantity $Q$ to be maximized or minimized and the variables involved. 2. Write a constraint relating the variables and use it to express $Q$ as a function of a single variable. 3. Determine the feasible domain. 4. Find the critical points and compare with the relevant endpoints or boundary behavior to locate the optimum.

步骤。1. 辨认要最大化或最小化的量 $Q$ 以及涉及的变量。2. 写出联系各变量的约束,并用它把 $Q$ 表示成单一变量的函数。3. 确定可行定义域。4. 求临界点,并与相关端点或边界行为比较,从而定位最优值。

Single-variable model单变量模型
$$ Q = Q(x), \quad x \in [\,x_{\min}, x_{\max}\,], \qquad \text{optimize over critical points and boundary.} $$
Worked Example 7.1: largest rectangle under a constraint例题 7.1:约束下面积最大的矩形

A farmer has $100$ meters of fencing to enclose a rectangular field. What dimensions maximize the area?

Let the rectangle have width $x$ and height $y$. The perimeter constraint is $2x + 2y = 100$, so $y = 50 - x$. The area to maximize is

$$ A(x) = x(50 - x) = 50x - x^{2}, \qquad x \in [0, 50]. $$

Then $A'(x) = 50 - 2x$, which vanishes at $x = 25$. Comparing candidates, $A(0) = 0$, $A(50) = 0$, and $A(25) = 25 \cdot 25 = 625$. The maximum area is $625$ square meters, achieved by the square with $x = y = 25$.

一位农夫有 $100$ 米篱笆,要围出一块矩形田地。什么尺寸能使面积最大?

设矩形宽 $x$、高 $y$。周长约束为 $2x + 2y = 100$,故 $y = 50 - x$。要最大化的面积为

$$ A(x) = x(50 - x) = 50x - x^{2}, \qquad x \in [0, 50]. $$

则 $A'(x) = 50 - 2x$,在 $x = 25$ 处为零。比较各候选:$A(0) = 0$、$A(50) = 0$、$A(25) = 25 \cdot 25 = 625$。最大面积为 $625$ 平方米,由 $x = y = 25$ 的正方形实现。

Going deeper: the cheapest cylindrical can深入一步:最省料的圆柱形罐头

Minimize the surface area of a closed cylindrical can of fixed volume $V$. With radius $r$ and height $h$, the volume constraint is $V = \pi r^{2} h$, so $h = V/(\pi r^{2})$. The surface area is

$$ S(r) = 2\pi r^{2} + 2\pi r h = 2\pi r^{2} + \frac{2V}{r}, \qquad r > 0. $$

Differentiate and set to zero.

$$ S'(r) = 4\pi r - \frac{2V}{r^{2}} = 0 \implies 4\pi r^{3} = 2V \implies r^{3} = \frac{V}{2\pi}. $$

Since $S''(r) = 4\pi + 4V/r^{3} > 0$ for all $r > 0$, this critical point is the absolute minimum. Substituting back, $h = V/(\pi r^{2}) = 2r$: the most economical can has height equal to its diameter.

使固定体积 $V$ 的密闭圆柱形罐头表面积最小。设半径 $r$、高 $h$,体积约束为 $V = \pi r^{2} h$,故 $h = V/(\pi r^{2})$。表面积为

$$ S(r) = 2\pi r^{2} + 2\pi r h = 2\pi r^{2} + \frac{2V}{r}, \qquad r > 0. $$

求导并令其为零。

$$ S'(r) = 4\pi r - \frac{2V}{r^{2}} = 0 \implies 4\pi r^{3} = 2V \implies r^{3} = \frac{V}{2\pi}. $$

由于对一切 $r > 0$ 有 $S''(r) = 4\pi + 4V/r^{3} > 0$,这个临界点是绝对最小值。回代得 $h = V/(\pi r^{2}) = 2r$:最省料的罐头其高等于其直径。

Worked Example 7.2: shortest distance from a point to a parabola例题 7.2:点到抛物线的最短距离

Find the point on the parabola $y = x^{2}$ closest to the point $(0, \tfrac{3}{2})$.

A point on the curve is $(x, x^{2})$. Rather than minimize the distance, minimize the squared distance, which has the same minimizer and avoids the square root.

$$ D(x) = x^{2} + \left(x^{2} - \tfrac{3}{2}\right)^{2}, \qquad x \in \mathbb{R}. $$

Differentiate and simplify.

$$ D'(x) = 2x + 2\left(x^{2} - \tfrac{3}{2}\right)(2x) = 2x\left(1 + 2x^{2} - 3\right) = 2x\,(2x^{2} - 2) = 4x(x^{2} - 1). $$

The critical points are $x = 0$ and $x = \pm 1$. Since $D''(x) = 12x^2 - 4$, we have $D''(0) = -4 < 0$ (a local max of squared distance) and $D''(\pm 1) = 8 > 0$ (local minima). Comparing $D(0) = \tfrac{9}{4}$ with $D(\pm 1) = 1 + \tfrac14 = \tfrac54$, the minimum distance occurs at $x = \pm 1$. The two closest points are $(1,1)$ and $(-1,1)$, each at distance $\sqrt{5}/2$. Minimizing the square is a standard device worth remembering.

求抛物线 $y = x^{2}$ 上离点 $(0, \tfrac{3}{2})$ 最近的点。

曲线上的点为 $(x, x^{2})$。与其最小化距离,不如最小化距离的平方,二者最小值点相同,且可避开平方根。

$$ D(x) = x^{2} + \left(x^{2} - \tfrac{3}{2}\right)^{2}, \qquad x \in \mathbb{R}. $$

求导并化简。

$$ D'(x) = 2x + 2\left(x^{2} - \tfrac{3}{2}\right)(2x) = 2x\left(1 + 2x^{2} - 3\right) = 2x\,(2x^{2} - 2) = 4x(x^{2} - 1). $$

临界点为 $x = 0$ 与 $x = \pm 1$。由 $D''(x) = 12x^2 - 4$,有 $D''(0) = -4 < 0$(平方距离的局部最大值)与 $D''(\pm 1) = 8 > 0$(局部最小值)。比较 $D(0) = \tfrac{9}{4}$ 与 $D(\pm 1) = 1 + \tfrac14 = \tfrac54$,最短距离在 $x = \pm 1$ 处。两个最近点是 $(1,1)$ 与 $(-1,1)$,各距 $\sqrt{5}/2$。最小化平方是值得记住的标准手法。

Worked Example 7.3: maximizing the volume of an open box例题 7.3:无盖盒子体积最大化

An open-top box is made from a $12 \times 12$ square sheet by cutting a square of side $x$ from each corner and folding up the sides. Find $x$ that maximizes the volume.

After folding, the base measures $(12 - 2x)$ by $(12 - 2x)$ and the height is $x$, so

$$ V(x) = x(12 - 2x)^{2}, \qquad x \in [0, 6]. $$

The physical domain is $[0,6]$ because the cut cannot exceed half the side. Differentiate using the product rule.

$$ V'(x) = (12 - 2x)^{2} + x\cdot 2(12 - 2x)(-2) = (12 - 2x)\big[(12 - 2x) - 4x\big] = (12 - 2x)(12 - 6x). $$

Setting $V'(x) = 0$ gives $x = 6$ or $x = 2$. The endpoints $x = 0$ and $x = 6$ both yield $V = 0$ (a flat sheet or a sealed-up box of zero base), so compare the interior candidate: $V(2) = 2(8)^{2} = 128$. The maximum volume is $128$ cubic units at $x = 2$. The boundary check is essential, since the algebra alone offers $x = 6$ as a spurious critical value sitting at the useless endpoint.

用一张 $12 \times 12$ 的方形纸板,在四角各剪去边长为 $x$ 的小方块再折起四边,做成一个无盖盒子。求使体积最大的 $x$。

折起后,底面为 $(12 - 2x)$ 乘 $(12 - 2x)$,高为 $x$,故

$$ V(x) = x(12 - 2x)^{2}, \qquad x \in [0, 6]. $$

实际定义域是 $[0,6]$,因为剪口不能超过边长的一半。用乘积法则求导。

$$ V'(x) = (12 - 2x)^{2} + x\cdot 2(12 - 2x)(-2) = (12 - 2x)\big[(12 - 2x) - 4x\big] = (12 - 2x)(12 - 6x). $$

令 $V'(x) = 0$ 得 $x = 6$ 或 $x = 2$。端点 $x = 0$ 与 $x = 6$ 都给出 $V = 0$(一张平板,或底面为零的封死盒子),故比较内部候选:$V(2) = 2(8)^{2} = 128$。最大体积为 $128$ 立方单位,在 $x = 2$ 处取得。边界检查至关重要,因为单凭代数会给出 $x = 6$ 这个落在无用端点上的虚假临界值。

Going deeper: the optimal angle in Snell's law of refraction深入一步:折射的斯涅尔定律中的最优角度

A classic application of optimization is to derive the law that governs how light bends. Suppose light travels from a point $A$ in one medium to a point $B$ in another, crossing a flat boundary, and that its speed is $v_1$ above the boundary and $v_2$ below. Fermat's principle of least time says the actual path is the one that minimizes travel time.

Place the boundary on the $x$-axis. Let the horizontal distance between $A$ and $B$ be $d$, with the crossing point at horizontal position $x$ measured from the foot of $A$. If $A$ is at height $a$ and $B$ at depth $b$, the total time is

$$ T(x) = \frac{\sqrt{a^{2} + x^{2}}}{v_1} + \frac{\sqrt{b^{2} + (d - x)^{2}}}{v_2}. $$

Differentiate and set $T'(x) = 0$:

$$ T'(x) = \frac{x}{v_1\sqrt{a^{2}+x^{2}}} - \frac{d - x}{v_2\sqrt{b^{2}+(d-x)^{2}}} = 0. $$

Now recognize the two fractions as sines of the angles each ray makes with the vertical (the normal to the boundary): if $\theta_1$ and $\theta_2$ are those angles of incidence and refraction, then $\sin\theta_1 = \dfrac{x}{\sqrt{a^{2}+x^{2}}}$ and $\sin\theta_2 = \dfrac{d-x}{\sqrt{b^{2}+(d-x)^{2}}}$. The stationarity condition becomes

$$ \frac{\sin\theta_1}{v_1} = \frac{\sin\theta_2}{v_2}, $$

which is exactly Snell's law. Because $T(x) \to \infty$ as $x \to \pm\infty$ and $T$ is smooth with a single critical point, that critical point is the global minimum. A purely geometric law of physics falls out of a one-variable minimization, a striking illustration of why optimization is a centerpiece of first-year calculus.

最优化的一个经典应用是推导支配光线弯折的定律。设光线从一种介质中的点 $A$ 出发,穿过一条平直界面,到达另一种介质中的点 $B$,且其速率在界面上方为 $v_1$、下方为 $v_2$。费马的最短时间原理指出,真实路径是使传播时间最短的那条。

把界面放在 $x$ 轴上。设 $A$ 与 $B$ 的水平距离为 $d$,过界点的水平位置为 $x$(从 $A$ 的垂足起算)。若 $A$ 在高度 $a$、$B$ 在深度 $b$,则总时间为

$$ T(x) = \frac{\sqrt{a^{2} + x^{2}}}{v_1} + \frac{\sqrt{b^{2} + (d - x)^{2}}}{v_2}. $$

求导并令 $T'(x) = 0$:

$$ T'(x) = \frac{x}{v_1\sqrt{a^{2}+x^{2}}} - \frac{d - x}{v_2\sqrt{b^{2}+(d-x)^{2}}} = 0. $$

现把这两个分式认作每条光线与竖直方向(界面法线)夹角的正弦:若入射角与折射角分别为 $\theta_1$ 与 $\theta_2$,则 $\sin\theta_1 = \dfrac{x}{\sqrt{a^{2}+x^{2}}}$,$\sin\theta_2 = \dfrac{d-x}{\sqrt{b^{2}+(d-x)^{2}}}$。驻点条件化为

$$ \frac{\sin\theta_1}{v_1} = \frac{\sin\theta_2}{v_2}, $$

这正是斯涅尔定律(Snell's law)。由于 $x \to \pm\infty$ 时 $T(x) \to \infty$,且 $T$ 光滑、只有一个临界点,该临界点就是全局最小值。一条纯几何的物理定律竟从一元最小化中落出,鲜明地说明了为何最优化是大学一年级微积分的核心。

Common error. The most damaging mistake in applied optimization is skipping the boundary and domain analysis. Finding $f'(x) = 0$ locates only interior critical points; if the feasible domain is a closed interval, the endpoints must be compared too, and a critical point outside the physical domain must be discarded. A second common slip is forgetting to use the constraint to reduce to a single variable before differentiating. Finally, always answer the question that was asked: if the problem wants the dimensions, do not stop at the critical value of the intermediate variable.
常见错误。应用最优化中最致命的错误是跳过边界与定义域分析。解 $f'(x) = 0$ 只定位内部临界点;若可行定义域是闭区间,还必须比较端点,而落在实际定义域之外的临界点必须舍去。第二个常见疏漏是求导前忘了用约束化简到单一变量。最后,始终回答题目所问:若问的是尺寸,就别停在中间变量的临界值上。
A rectangular pen against a wall uses fencing on three sides only, with total fencing $L$. To maximize area, the side parallel to the wall should equal一个靠墙的矩形围栏只在三面用篱笆,篱笆总长为 $L$。要使面积最大,与墙平行的边应等于
7.1
$L/4$$L/4$
$L/2$$L/2$
$L/3$$L/3$
$2L/3$$2L/3$
Correct. With parallel side $y$ and two perpendicular sides $x$, the constraint is $2x + y = L$, so area $A = x(L-2x)$. Then $A'=L-4x=0$ gives $x=L/4$ and $y=L-2x=L/2$. The wall-parallel side is $L/2$.正确。设平行边为 $y$、两条垂直边各为 $x$,约束为 $2x + y = L$,故面积 $A = x(L-2x)$。由 $A'=L-4x=0$ 得 $x=L/4$,$y=L-2x=L/2$。与墙平行的边为 $L/2$。
Model it as $2x + y = L$ with area $x y = x(L-2x)$. Maximizing gives $x = L/4$, hence the parallel side $y = L - 2x = L/2$.建模为 $2x + y = L$,面积 $x y = x(L-2x)$。最大化得 $x = L/4$,故平行边 $y = L - 2x = L/2$。

Flashcards速记卡

0 / 12 flipped已翻 0 / 12
Critical point of $f$$f$ 的临界点
A number $c$ in the domain where $f'(c)=0$ or $f'(c)$ does not exist. These are the only interior candidates for local extrema.定义域中满足 $f'(c)=0$ 或 $f'(c)$ 不存在的数 $c$。它们是局部极值仅有的内部候选。
Fermat's Theorem费马定理
If $f$ has a local extremum at $c$ and $f'(c)$ exists, then $f'(c)=0$. The converse is false: $f(x)=x^{3}$ has $f'(0)=0$ but no extremum.若 $f$ 在 $c$ 处有局部极值且 $f'(c)$ 存在,则 $f'(c)=0$。逆命题不成立:$f(x)=x^{3}$ 有 $f'(0)=0$ 却无极值。
Extreme Value Theorem极值定理
A continuous function on a closed bounded interval $[a,b]$ attains both an absolute maximum and an absolute minimum on $[a,b]$.在有界闭区间 $[a,b]$ 上连续的函数,必在 $[a,b]$ 上同时取得绝对最大值与绝对最小值。
Mean Value Theorem中值定理
If $f$ is continuous on $[a,b]$ and differentiable on $(a,b)$, some $c\in(a,b)$ satisfies 若 $f$ 在 $[a,b]$ 上连续、在 $(a,b)$ 上可微,则存在某个 $c\in(a,b)$ 满足 $$f'(c)=\frac{f(b)-f(a)}{b-a}.$$
Rolle's Theorem罗尔定理
If $f$ is continuous on $[a,b]$, differentiable on $(a,b)$, and $f(a)=f(b)$, then $f'(c)=0$ for some $c\in(a,b)$.若 $f$ 在 $[a,b]$ 上连续、在 $(a,b)$ 上可微,且 $f(a)=f(b)$,则存在某个 $c\in(a,b)$ 使 $f'(c)=0$。
Increasing / Decreasing Test单调性判别法
If $f'>0$ on an interval then $f$ increases there; if $f'<0$ then $f$ decreases there. Proved via the Mean Value Theorem.若在某区间上 $f'>0$,则 $f$ 在该区间递增;若 $f'<0$,则递减。由中值定理证明。
First Derivative Test一阶导数判别法
At a critical point $c$: $f'$ from $(+)$ to $(-)$ gives a local max, $(-)$ to $(+)$ gives a local min, and no sign change gives neither.在临界点 $c$:$f'$ 由 $(+)$ 变 $(-)$ 给出局部最大,$(-)$ 变 $(+)$ 给出局部最小,无符号变化则皆非。
Concavity Test凹凸性判别法
If $f''>0$ on an interval, $f$ is concave up; if $f''<0$, concave down. An inflection point is where concavity reverses.若在某区间上 $f''>0$,则 $f$ 凹向上;若 $f''<0$,则凹向下。拐点是凹凸性反转之处。
Second Derivative Test二阶导数判别法
If $f'(c)=0$: $f''(c)>0$ gives a local minimum and $f''(c)<0$ gives a local maximum. If $f''(c)=0$ the test is inconclusive.若 $f'(c)=0$:$f''(c)>0$ 给出局部最小,$f''(c)<0$ 给出局部最大。若 $f''(c)=0$ 则无法判定。
Closed Interval Method闭区间法
For $f$ continuous on $[a,b]$, evaluate $f$ at all critical points in $(a,b)$ and at $a$ and $b$; the largest value is the absolute max and the smallest is the absolute min.对在 $[a,b]$ 上连续的 $f$,在 $(a,b)$ 内所有临界点以及 $a$、$b$ 处求 $f$ 的值;最大的是绝对最大值,最小的是绝对最小值。
Horizontal asymptote水平渐近线
If $\lim_{x\to\pm\infty} f(x)=L$ then $y=L$ is a horizontal asymptote. For a rational function with equal degrees, $L$ is the ratio of leading coefficients.若 $\lim_{x\to\pm\infty} f(x)=L$,则 $y=L$ 是水平渐近线。对分子分母次数相同的有理函数,$L$ 是首项系数之比。
Applied optimization recipe应用最优化套路
Name the quantity $Q$, use a constraint to write $Q=Q(x)$ in one variable, fix the feasible domain, then compare critical points with the boundary.给量 $Q$ 命名,用约束把 $Q=Q(x)$ 写成单变量,定下可行定义域,再把临界点与边界比较。

Unit Quiz单元测验

The critical points of $f(x) = x^{3} - 12x$ are$f(x) = x^{3} - 12x$ 的临界点是
Q1
$x = 0$ only只有 $x = 0$
$x = \pm 12$$x = \pm 12$
$x = \pm 2$$x = \pm 2$
there are none没有临界点
Correct. $f'(x) = 3x^{2} - 12 = 3(x^{2}-4)$ vanishes at $x = \pm 2$, and $f'$ exists everywhere, so these are the only critical points.正确。$f'(x) = 3x^{2} - 12 = 3(x^{2}-4)$ 在 $x = \pm 2$ 处为零,且 $f'$ 处处存在,故这就是仅有的临界点。
Solve $f'(x) = 3x^{2} - 12 = 0$, giving $x^{2} = 4$ and $x = \pm 2$.解 $f'(x) = 3x^{2} - 12 = 0$,得 $x^{2} = 4$,$x = \pm 2$。
Which hypothesis of the Mean Value Theorem is missing if $f$ is differentiable on $(a,b)$ but only known to be continuous on $(a,b)$?若 $f$ 在 $(a,b)$ 上可微,但仅知在 $(a,b)$ 上连续,那么中值定理缺了哪个前提?
Q2
Continuity on the closed interval $[a,b]$在闭区间 $[a,b]$ 上连续
Differentiability at $a$ and $b$在 $a$ 与 $b$ 处可微
That $f(a) = f(b)$$f(a) = f(b)$
Nothing is missing没有缺任何前提
Correct. The Mean Value Theorem requires continuity on the closed interval $[a,b]$, including the endpoints, in addition to differentiability on the open interval.正确。中值定理除了要求在开区间上可微,还要求在闭区间 $[a,b]$ 上(含端点)连续。
The theorem needs continuity on all of $[a,b]$, endpoints included. It does not require differentiability at the endpoints or that $f(a)=f(b)$.定理需要在整个 $[a,b]$(含端点)上连续。它不要求在端点处可微,也不要求 $f(a)=f(b)$。
If $f'(x) > 0$ on $(1,4)$ and $f'(x) < 0$ on $(4,7)$, then at $x = 4$ the function $f$ has a若 $f'(x) > 0$ 在 $(1,4)$ 上、$f'(x) < 0$ 在 $(4,7)$ 上,则在 $x = 4$ 处函数 $f$ 有一个
Q3
local minimum局部最小值
inflection point拐点
vertical asymptote竖直渐近线
local maximum局部最大值
Correct. The first derivative changes from positive to negative across $x = 4$, so by the First Derivative Test $f$ has a local maximum there.正确。一阶导数在 $x = 4$ 处由正变负,故由一阶导数判别法,$f$ 在此有局部最大值。
A $(+)\to(-)$ change in $f'$ marks a local maximum: the function rises, then falls.$f'$ 的 $(+)\to(-)$ 变化标志局部最大值:函数先升后降。
The function $f(x) = x^{3}$ has an inflection point at $x = 0$ because函数 $f(x) = x^{3}$ 在 $x = 0$ 处有拐点,原因是
Q4
$f'(0) = 0$$f'(0) = 0$
$f''(x) = 6x$ changes sign at $0$$f''(x) = 6x$ 在 $0$ 处变号
$f$ has a local minimum there$f$ 在此有局部最小值
$f''(0) \ne 0$$f''(0) \ne 0$
Correct. An inflection point requires the concavity to reverse. Here $f''(x) = 6x$ is negative for $x<0$ and positive for $x>0$, so the sign changes at $0$.正确。拐点要求凹凸性反转。这里 $f''(x) = 6x$ 在 $x<0$ 时为负、$x>0$ 时为正,故在 $0$ 处变号。
An inflection point is about a change in concavity, that is, a sign change in $f''$. For $x^{3}$, $f''=6x$ changes sign at $0$.拐点关乎凹凸性的改变,即 $f''$ 变号。对 $x^{3}$,$f''=6x$ 在 $0$ 处变号。
The absolute maximum of $f(x) = x^{2} - 4x + 1$ on $[0, 3]$ is attained at$f(x) = x^{2} - 4x + 1$ 在 $[0, 3]$ 上的绝对最大值在何处取得
Q5
$x = 2$, the critical point$x = 2$,即临界点
$x = 3$$x = 3$
$x = 0$$x = 0$
$x = 1$$x = 1$
Correct. $f'(x) = 2x - 4$ gives the critical point $x = 2$. Evaluating candidates: $f(0) = 1$, $f(2) = -3$, $f(3) = -2$. The largest is $f(0) = 1$, so the maximum is at the endpoint $x = 0$.正确。$f'(x) = 2x - 4$ 给出临界点 $x = 2$。求各候选:$f(0) = 1$、$f(2) = -3$、$f(3) = -2$。最大的是 $f(0) = 1$,故最大值在端点 $x = 0$ 处。
Compare $f$ at the critical point and both endpoints: $f(0)=1$, $f(2)=-3$, $f(3)=-2$. The maximum value $1$ occurs at $x=0$.在临界点与两个端点比较 $f$:$f(0)=1$、$f(2)=-3$、$f(3)=-2$。最大值 $1$ 出现在 $x=0$。
To enclose a rectangle of area $A=900$ with the least perimeter, the rectangle should be要以最小周长围出面积 $A=900$ 的矩形,该矩形应为
Q6
a square with side $30$边长 $30$ 的正方形
$90$ by $10$$90$ 乘 $10$
$45$ by $20$$45$ 乘 $20$
$180$ by $5$$180$ 乘 $5$
Correct. With $xy=900$, the perimeter $P=2x+2(900/x)$ has $P'=2-1800/x^{2}=0$ at $x^{2}=900$, so $x=30$ and $y=30$. The optimal rectangle is the square of side $30$.正确。由 $xy=900$,周长 $P=2x+2(900/x)$ 满足 $P'=2-1800/x^{2}=0$,即 $x^{2}=900$,故 $x=30$、$y=30$。最优矩形是边长 $30$ 的正方形。
Minimizing $P = 2x + 1800/x$ gives $x^{2} = 900$, so $x = y = 30$: the square minimizes perimeter for fixed area.最小化 $P = 2x + 1800/x$ 得 $x^{2} = 900$,故 $x = y = 30$:面积固定时正方形使周长最小。

Readiness Checklist掌握度清单

Tap each item you can do without notes.点选每一项你能不看笔记就完成的内容。 0 / 8 mastered已掌握 0 / 8