索引
微分中值定理
极值
定义4.1 极大(小)值
若存在
x
0
x_{0}
x0的邻域
U
(
x
0
,
δ
)
U\left ( x_{0}, \delta \right )
U(x0,δ),使得
∀
x
∈
U
(
x
0
)
,
δ
\forall x\in U\left ( x_{0}\right), \delta
∀x∈U(x0),δ,
f
(
x
)
≤
f
(
x
0
)
f\left(x\right ) \le f\left(x_{0}\right )
f(x)≤f(x0),则称
x
0
x_{0}
x0是函数
f
(
x
)
f\left(x\right )
f(x)的一个极大值点。
若存在
x
0
x_{0}
x0的邻域
U
(
x
0
,
δ
)
U\left ( x_{0}, \delta \right )
U(x0,δ),使得
∀
x
∈
U
(
x
0
)
,
δ
\forall x\in U\left ( x_{0}\right), \delta
∀x∈U(x0),δ,
f
(
x
)
≥
f
(
x
0
)
f\left(x\right ) \ge f\left(x_{0}\right )
f(x)≥f(x0),则称
x
0
x_{0}
x0是函数
f
(
x
)
f\left(x\right )
f(x)的一个极小值点。
定理4.1 Fermat引理
设 x 0 x_{0} x0是函数 f ( x ) f\left(x\right ) f(x)的一个极值点,且函数 f ( x ) f\left(x\right ) f(x)在点 x 0 x_{0} x0处可导,则 f ′ ( x 0 ) = 0 f^{\prime }\left ( x_{0} \right )=0 f′(x0)=0。
不妨设
x
0
x_{0}
x0是函数
f
(
x
)
f\left(x\right )
f(x)的一个极大值点。
令
x
∈
(
x
0
,
x
0
+
δ
)
x\in \left ( x_{0},x_{0}+\delta \right )
x∈(x0,x0+δ),则
f
(
x
0
+
Δ
x
)
−
f
(
x
0
)
Δ
x
≤
0
\frac{f\left ( x_{0}+\Delta x \right )-f\left ( x_{0} \right ) }{\Delta x} \le 0
Δxf(x0+Δx)−f(x0)≤0,
根据函数极限的保不等号性,
f
+
′
(
x
0
)
=
lim
x
0
+
→
0
f
(
x
0
+
Δ
x
)
−
f
(
x
0
)
Δ
x
≤
0
f^{\prime } _{+}\left ( x_{0} \right ) =\lim_{x_{0}^{+} \to 0} \frac{f\left ( x_{0}+\Delta x \right )-f\left ( x_{0} \right ) }{\Delta x}\le 0
f+′(x0)=limx0+→0Δxf(x0+Δx)−f(x0)≤0;
同理令
x
∈
(
x
0
−
δ
,
x
0
)
x\in \left ( x_{0}-\delta,x_{0}\right )
x∈(x0−δ,x0),则
f
−
′
(
x
0
)
=
lim
x
0
−
→
0
f
(
x
0
+
Δ
x
)
−
f
(
x
0
)
Δ
x
≥
0
f^{\prime } _{-}\left ( x_{0} \right ) =\lim_{x_{0}^{-} \to 0} \frac{f\left ( x_{0}+\Delta x \right )-f\left ( x_{0} \right ) }{\Delta x}\ge 0
f−′(x0)=limx0−→0Δxf(x0+Δx)−f(x0)≥0,
因为函数
f
(
x
)
f\left(x\right )
f(x)在点
x
0
x_{0}
x0处可导,所以函数
f
(
x
)
f\left(x\right )
f(x)在点
x
0
x_{0}
x0处
f
+
′
(
x
0
)
=
f
−
′
(
x
0
)
=
f
′
(
x
0
)
=
0
f^{\prime } _{+}\left ( x_{0} \right ) =f^{\prime } _{-}\left ( x_{0} \right ) =f^{\prime }\left ( x_{0} \right )=0
f+′(x0)=f−′(x0)=f′(x0)=0。
定理4.2 Rolle定理
函数 f ( x ) f\left ( x \right ) f(x)在闭区间 [ a , b ] \left [ a,b \right ] [a,b]上连续,在开区间 ( a , b ) \left ( a,b \right ) (a,b)上可导, f ( a ) = f ( b ) f\left ( a \right )=f\left ( b \right ) f(a)=f(b),则 ∃ ξ ∈ ( a , b ) \exists \xi \in \left ( a,b \right ) ∃ξ∈(a,b): f ′ ( ξ ) = 0 f^{\prime }\left ( \xi \right )=0 f′(ξ)=0。
根据最值定理,
f
(
x
)
f\left ( x \right )
f(x)在
[
a
,
b
]
\left [ a,b \right ]
[a,b]上必有最大值
M
M
M和最小值
m
m
m,也就是
∃
η
\exists \eta
∃η,
ξ
∈
[
a
,
b
]
\xi \in \left [ a,b \right ]
ξ∈[a,b]:
∀
x
∈
[
a
,
b
]
\forall x\in \left [ a,b \right ]
∀x∈[a,b],
f
(
η
)
=
m
=
min
f
(
x
)
f\left ( \eta \right )=m= \min f\left ( x \right )
f(η)=m=minf(x),
f
(
ξ
)
=
M
=
max
f
(
x
)
f\left ( \xi \right )=M=\max f\left ( x \right )
f(ξ)=M=maxf(x)。
<1>
M
=
m
M=m
M=m
此时函数
f
(
x
)
f\left ( x \right )
f(x)为常数函数,显然
∃
ξ
∈
(
a
,
b
)
\exists \xi \in \left ( a,b \right )
∃ξ∈(a,b):
f
′
(
ξ
)
=
0
f^{\prime }\left ( \xi \right )=0
f′(ξ)=0。
<2>
M
>
m
M> m
M>m
此时
M
=
f
(
ξ
)
M=f\left ( \xi \right )
M=f(ξ)为
f
(
x
)
f\left ( x \right )
f(x)在
[
a
,
b
]
\left [ a,b \right ]
[a,b]上的一个极大值,
ξ
\xi
ξ是函数
f
(
x
)
f\left(x\right )
f(x)的一个极大值点,
根据Fermat引理,
f
′
(
ξ
)
=
0
f^{\prime }\left ( \xi \right )=0
f′(ξ)=0。
Lagrange中值定理
定理4.3 Lagrange中值定理
函数 y = f ( x ) y=f\left ( x \right ) y=f(x)在闭区间 [ a , b ] \left [ a,b \right ] [a,b]上连续,在开区间 ( a , b ) \left ( a,b \right ) (a,b)上可导,则 ∃ ξ ∈ ( a , b ) \exists \xi \in \left ( a,b \right ) ∃ξ∈(a,b): f ′ ( ξ ) = f ( b ) − f ( a ) b − a f^{\prime } \left ( \xi \right )=\frac{f\left ( b\right ) -f\left ( a \right ) }{b-a} f′(ξ)=b−af(b)−f(a)。
任取
x
∈
(
a
,
b
)
x \in \left ( a,b \right )
x∈(a,b),
x
=
t
x=t
x=t处切线斜率为
f
′
(
t
)
f^{\prime } \left ( t \right )
f′(t)。
另外连接闭区间
[
a
,
b
]
\left [ a,b \right ]
[a,b]端点的割线斜率为
k
=
f
(
b
)
−
f
(
a
)
b
−
a
k=\frac{f\left ( b \right )-f\left ( a \right ) }{b-a}
k=b−af(b)−f(a),割线方程为
y
−
f
(
a
)
=
(
f
(
b
)
−
f
(
a
)
b
−
a
)
(
x
−
a
)
y-f\left ( a \right )=\left ( \frac{f\left ( b \right )-f\left ( a \right ) }{b-a} \right )\left ( x-a \right )
y−f(a)=(b−af(b)−f(a))(x−a),
而点
(
t
,
f
(
t
)
)
\left (t ,f\left (t \right ) \right )
(t,f(t))到割线
y
=
(
f
(
b
)
−
f
(
a
)
b
−
a
)
(
x
−
a
)
y=\left ( \frac{f\left ( b \right )-f\left ( a \right ) }{b-a} \right )\left ( x-a \right )
y=(b−af(b)−f(a))(x−a)的距离函数为
d
(
x
)
=
∣
k
(
x
−
a
)
+
f
(
a
)
−
f
(
x
)
∣
1
+
k
2
d\left ( x \right )=\frac{\left | k\left ( x-a \right )+f\left ( a \right ) -f\left ( x \right )\right | }{\sqrt{1+k^{2} } }
d(x)=1+k2∣k(x−a)+f(a)−f(x)∣,
d
′
(
x
)
=
∣
k
−
f
′
(
x
)
∣
1
+
k
2
d^{\prime } \left ( x \right ) =\frac{\left | k-f^{\prime } \left ( x \right ) \right | }{\sqrt{1+k^{2} } }
d′(x)=1+k2∣k−f′(x)∣,
因为
d
(
x
)
d\left ( x \right )
d(x)在闭区间
[
a
,
b
]
\left [ a,b \right ]
[a,b]上连续,在开区间
(
a
,
b
)
\left ( a,b \right )
(a,b)上可导,且
d
(
a
)
=
d
(
b
)
=
0
d \left ( a \right )= d \left ( b \right )=0
d(a)=d(b)=0,所以根据Rolle定理,
∃
ξ
∈
(
a
,
b
)
\exists \xi \in \left ( a,b \right )
∃ξ∈(a,b):
d
′
(
ξ
)
=
0
d^{\prime }\left ( \xi \right )=0
d′(ξ)=0,
解方程
d
′
(
ξ
)
=
∣
k
−
f
′
(
ξ
)
∣
1
+
k
2
=
0
d^{\prime }\left ( \xi \right )=\frac{\left | k-f^{\prime } \left ( \xi \right ) \right | }{\sqrt{1+k^{2} } }=0
d′(ξ)=1+k2∣k−f′(ξ)∣=0,可得
f
′
(
ξ
)
=
k
=
f
(
b
)
−
f
(
a
)
b
−
a
f^{\prime } \left ( \xi \right ) =k=\frac{f\left ( b \right )-f\left ( a \right ) }{b-a}
f′(ξ)=k=b−af(b)−f(a)。
从几何意义出发,同样根据距离函数
d
(
t
)
d\left ( t \right )
d(t)的分子部分,可以构造函数
φ
(
x
)
=
f
(
x
)
−
f
(
a
)
−
f
(
b
)
−
f
(
a
)
b
−
a
(
x
−
a
)
=
f
(
x
)
−
f
(
a
)
−
k
(
x
−
a
)
\varphi \left ( x \right )=f\left ( x \right )-f\left ( a \right )-\frac{f\left ( b \right )-f\left ( a \right ) }{b-a}\left ( x-a \right )= f\left ( x \right )-f\left ( a \right )-k\left ( x-a \right )
φ(x)=f(x)−f(a)−b−af(b)−f(a)(x−a)=f(x)−f(a)−k(x−a),
φ
(
x
)
\varphi \left ( x \right )
φ(x)仍然满足Rolle定理,
∃
ξ
∈
(
a
,
b
)
\exists \xi \in \left ( a,b \right )
∃ξ∈(a,b):
φ
′
(
ξ
)
=
0
\varphi^{\prime } \left ( \xi \right )=0
φ′(ξ)=0,代数方法与几何方法实质上殊途同归。
定理4.4 Cauchy中值定理
函数 y = f ( x ) y=f\left ( x \right ) y=f(x)在闭区间 [ a , b ] \left [ a,b \right ] [a,b]上连续,在开区间 ( a , b ) \left ( a,b \right ) (a,b)上可导,则 ∃ ξ ∈ ( a , b ) \exists \xi \in \left ( a,b \right ) ∃ξ∈(a,b): f ′ ( ξ ) g ′ ( ξ ) = f ( b ) − f ( a ) g ( b ) − g ( a ) ( g ′ ( ξ ) ≠ 0 ) \frac{f^{\prime }\left ( \xi \right ) }{g^{\prime }\left ( \xi \right ) }=\frac{f\left ( b \right )-f\left ( a \right ) }{g\left ( b \right )-g\left ( a \right ) }\left ( g^{\prime}\left ( \xi \right ) \neq 0 \right ) g′(ξ)f′(ξ)=g(b)−g(a)f(b)−f(a)(g′(ξ)=0)。
联立参数方程:
{
y
=
f
(
t
)
x
=
g
(
t
)
\left\{\begin{matrix} y=f\left ( t \right ) \\ x=g\left ( t \right ) \end{matrix}\right.
{y=f(t)x=g(t)
参考Lagrange中值定理,将距离函数中的纵坐标
y
y
y替换为参数方程
y
=
f
(
t
)
y=f\left ( t \right )
y=f(t),距离函数中的横坐标
x
x
x替换为参数方程
x
=
g
(
t
)
x=g\left ( t \right )
x=g(t),可构造函数
φ
(
t
)
=
f
(
t
)
−
f
(
a
)
−
(
f
(
b
)
−
f
(
a
)
g
(
b
)
−
g
(
a
)
)
(
g
(
t
)
−
g
(
a
)
)
\varphi \left ( t \right )=f\left ( t \right )-f\left ( a \right )-\left ( \frac{f\left ( b \right )-f\left ( a \right ) }{g\left ( b \right )-g\left ( a \right ) } \right ) \left ( g\left ( t \right ) -g\left ( a \right ) \right )
φ(t)=f(t)−f(a)−(g(b)−g(a)f(b)−f(a))(g(t)−g(a)),后续过程与Lagrange中值定理一致。
一阶导数与单调性的关系
定理4.5
函数
f
(
x
)
f\left ( x \right )
f(x)在区间
I
I
I上有定义,且一阶可导,则函数
f
(
x
)
f\left ( x \right )
f(x)在区间
I
I
I上单调增加(减少)的充要条件是
∀
x
∈
I
\forall x\in I
∀x∈I,
f
′
(
x
)
≥
0
(
f
′
(
x
)
≤
0
)
f^{\prime }\left ( x \right ) \ge 0(f^{\prime }\left ( x \right ) \le 0)
f′(x)≥0(f′(x)≤0)。
函数
f
(
x
)
f\left ( x \right )
f(x)在区间
I
I
I上严格单调增加(严格减少)的充要条件是
∀
x
∈
I
\forall x\in I
∀x∈I,
f
′
(
x
)
>
0
(
f
′
(
x
)
<
0
)
f^{\prime }\left ( x \right )>0(f^{\prime }\left ( x \right ) < 0)
f′(x)>0(f′(x)<0)。
<1>必要性
设
x
>
x
0
x>x_{0}
x>x0,
引用Lagrange定理,
∃
ξ
∈
I
\exists \xi \in I
∃ξ∈I,
f
(
x
)
−
f
(
x
0
)
=
f
′
(
ξ
)
(
x
−
x
0
)
f\left ( x \right )-f\left ( x_{0} \right ) =f^{\prime }\left ( \xi \right )\left ( x-x_{0} \right )
f(x)−f(x0)=f′(ξ)(x−x0),
又有
∀
x
∈
I
\forall x\in I
∀x∈I,
f
′
(
x
)
≥
0
f^{\prime }\left ( x \right ) \ge 0
f′(x)≥0,所以
f
′
(
ξ
)
≥
0
f^{\prime }\left ( \xi \right ) \ge 0
f′(ξ)≥0,
f
(
x
)
−
f
(
x
0
)
≥
0
f\left ( x \right )-f\left ( x_{0} \right ) \ge 0
f(x)−f(x0)≥0,
f
(
x
)
≥
f
(
x
0
)
f\left ( x \right ) \ge f\left ( x_{0} \right )
f(x)≥f(x0)。
<2>充分性
f
(
x
)
−
f
(
x
0
)
x
−
x
0
≥
0
\frac{f\left ( x \right )-f\left ( x_{0} \right )}{x-x_{0} }\ge 0
x−x0f(x)−f(x0)≥0,
引用函数极限的局部保号性,令
Δ
x
=
x
−
x
0
\Delta x=x-x_{0}
Δx=x−x0,
lim
Δ
x
→
0
f
(
x
)
−
f
(
x
0
)
x
−
x
0
≥
0
\lim _{\Delta x\to 0} \frac{f\left ( x \right )-f\left ( x_{0} \right )}{x-x_{0} }\ge 0
limΔx→0x−x0f(x)−f(x0)≥0,
根据导数的定义,
f
′
(
x
)
≥
0
f^{\prime} \left ( x \right ) \ge 0
f′(x)≥0。
二阶导数与凸性的关系
定义4.2 凸性
函数
f
(
x
)
f\left ( x \right )
f(x)在区间
I
I
I有定义,若
∀
x
1
\forall x_{1}
∀x1,
x
2
∈
I
x_{2}\in I
x2∈I,
∀
λ
∈
(
0
,
1
)
\forall \lambda \in \left ( 0,1 \right )
∀λ∈(0,1),有
f
(
λ
x
1
+
(
1
−
λ
)
x
2
)
≤
λ
f
(
x
1
)
+
(
1
−
λ
)
f
(
x
2
)
f\left ( \lambda x_{1} +\left ( 1-\lambda \right ) x_{2} \right )\le \lambda f\left ( x_{1} \right ) +\left ( 1-\lambda \right )f\left ( x_{2} \right )
f(λx1+(1−λ)x2)≤λf(x1)+(1−λ)f(x2),则称函数
f
(
x
)
f\left ( x \right )
f(x)是下凸函数,几何意义为连接区间端点的割线段恒在函数曲线上方。
若
∀
x
1
\forall x_{1}
∀x1,
x
2
∈
I
x_{2}\in I
x2∈I,
∀
λ
∈
(
0
,
1
)
\forall \lambda \in \left ( 0,1 \right )
∀λ∈(0,1),有
f
(
λ
x
1
+
(
1
−
λ
)
x
2
)
≥
λ
f
(
x
1
)
+
(
1
−
λ
)
f
(
x
2
)
f\left ( \lambda x_{1} +\left ( 1-\lambda \right ) x_{2} \right )\ge \lambda f\left ( x_{1} \right ) +\left ( 1-\lambda \right )f\left ( x_{2} \right )
f(λx1+(1−λ)x2)≥λf(x1)+(1−λ)f(x2),则称函数
f
(
x
)
f\left ( x \right )
f(x)是上凸函数,几何意义为连接区间端点的割线段恒在函数曲线上下方。
定理4.6
函数
f
(
x
)
f\left ( x \right )
f(x)在区间I上二阶可导,函数
f
(
x
)
f\left ( x \right )
f(x)在区间
I
I
I上下凸(上凸)的充要条件是
∀
x
∈
I
\forall x\in I
∀x∈I,
f
′
′
(
x
)
≥
0
f^{\prime \prime }\left ( x \right )\ge 0
f′′(x)≥0(
f
′
′
(
x
)
≤
0
f^{\prime \prime } \left ( x \right ) \le 0
f′′(x)≤0)。
函数
f
(
x
)
f\left ( x \right )
f(x)在区间
I
I
I上严格下凸(上凸)的充要条件是
∀
x
∈
I
\forall x\in I
∀x∈I,
f
′
′
(
x
)
>
0
f^{\prime \prime }\left ( x \right )> 0
f′′(x)>0(
f
′
′
(
x
)
<
0
f^{\prime \prime } \left ( x \right ) < 0
f′′(x)<0)。
不妨设函数
f
(
x
)
f\left ( x \right )
f(x)在区间
I
I
I上下凸,
∀
x
∈
I
\forall x\in I
∀x∈I,
f
′
′
(
x
)
≤
0
f^{\prime \prime }\left ( x \right )\le 0
f′′(x)≤0,
x
1
<
x
2
x_{1}<x_{2}
x1<x2,
<1>必要性
令
λ
=
1
2
\lambda =\frac{1}{2}
λ=21,
f
(
x
1
+
x
2
2
)
≤
f
(
x
1
)
+
f
(
x
2
)
2
⇒
f
(
x
2
)
−
f
(
x
1
+
x
2
2
)
≥
f
(
x
1
+
x
2
2
)
−
f
(
x
1
)
f\left ( \frac{x_{1}+x_{2}}{2} \right ) \le \frac{f\left ( x_{1} \right ) +f\left ( x_{2} \right ) }{2}\Rightarrow f\left ( x_{2} \right ) -f\left ( \frac{x_{1}+x_{2} }{2} \right )\ge f\left ( \frac{x_{1} +x_{2}}{2} \right )-f\left ( x_{1} \right )
f(2x1+x2)≤2f(x1)+f(x2)⇒f(x2)−f(2x1+x2)≥f(2x1+x2)−f(x1),
设
Δ
x
n
=
∣
x
2
−
x
1
∣
n
=
x
2
−
x
1
n
>
0
\Delta x_{n} =\frac{\left | x_{2}-x_{1} \right | }{n}=\frac{x_{2}-x_{1} }{n}>0
Δxn=n∣x2−x1∣=nx2−x1>0(
n
≥
3
n\ge3
n≥3),
可抽象出如下递推公式,其中
k
∈
N
+
k\in \mathbb{N}^{+}
k∈N+,
k
∈
[
2
,
n
]
k\in \left [ 2,n \right ]
k∈[2,n]:
f
(
x
2
)
−
f
(
x
2
−
Δ
x
n
)
≤
f
(
x
2
−
Δ
x
n
)
−
f
(
x
2
−
2
Δ
x
n
)
⋯
≤
f
(
x
2
−
(
k
−
2
)
Δ
x
n
)
−
f
(
x
2
−
(
k
−
1
)
Δ
x
n
)
≤
f
(
x
2
−
(
k
−
1
)
Δ
x
n
)
−
f
(
x
2
−
k
Δ
x
n
)
⋯
≤
f
(
x
2
−
(
n
−
1
)
Δ
x
n
)
−
f
(
x
2
−
n
Δ
x
n
)
=
f
(
x
1
+
Δ
x
n
)
−
f
(
x
1
)
\begin{array}{l} & f\left ( x_{2} \right )-f\left ( x_{2} - \Delta x_{n} \right ) \\ \le & f\left ( x_{2} -\Delta x_{n} \right ) -f\left ( x_{2} -2 \Delta x_{n} \right )\\ \cdots \\ \le & f\left ( x_{2}-\left ( k-2 \right )\Delta x_{n} \right )-f\left ( x_{2}- \left ( k-1 \right )\Delta x_{n} \right ) \\ \le & f\left ( x_{2}-\left ( k-1 \right )\Delta x_{n} \right )-f\left ( x_{2}- k\Delta x_{n} \right )\\ \cdots \\ \le & f\left ( x_{2}-\left ( n-1 \right )\Delta x_{n} \right )-f\left ( x_{2}-n\Delta x_{n} \right ) \\ =& f\left ( x_{1}+\Delta x_{n} \right )-f\left ( x_{1} \right ) \\ \end{array}
≤⋯≤≤⋯≤=f(x2)−f(x2−Δxn)f(x2−Δxn)−f(x2−2Δxn)f(x2−(k−2)Δxn)−f(x2−(k−1)Δxn)f(x2−(k−1)Δxn)−f(x2−kΔxn)f(x2−(n−1)Δxn)−f(x2−nΔxn)f(x1+Δxn)−f(x1)
综上所述,
f
(
x
2
)
−
f
(
x
2
−
Δ
x
n
)
≤
f
(
x
1
+
Δ
x
n
)
−
f
(
x
1
)
f\left ( x_{2} \right )-f\left ( x_{2} - \Delta x_{n} \right ) \le f\left ( x_{1}+\Delta x_{n} \right )-f\left ( x_{1} \right )
f(x2)−f(x2−Δxn)≤f(x1+Δxn)−f(x1),记为(
∗
\ast
∗),
对不等式(
∗
\ast
∗)两侧同除以
Δ
x
n
\Delta x_{n}
Δxn,
f
(
x
2
)
−
f
(
x
2
+
(
−
Δ
x
n
)
)
−
Δ
x
n
≥
f
(
x
1
+
Δ
x
n
)
−
f
(
x
1
)
Δ
x
n
\frac{f\left ( x_{2} \right ) -f\left ( x_{2} +\left ( -\Delta x_{n} \right ) \right ) }{-\Delta x_{n} }\ge \frac{f\left ( x_{1} +\Delta x_{n} \right ) -f\left ( x_{1} \right ) }{\Delta x_{n} }
−Δxnf(x2)−f(x2+(−Δxn))≥Δxnf(x1+Δxn)−f(x1),
引用函数极限的局部保号性,
lim
Δ
x
n
→
0
f
(
x
2
)
−
f
(
x
2
+
(
−
Δ
x
n
)
)
−
Δ
x
n
≥
lim
Δ
x
n
→
0
f
(
x
1
+
Δ
x
n
)
−
f
(
x
1
)
Δ
x
n
\lim _{\Delta x_{n}\to 0 } \frac{f\left ( x_{2} \right ) -f\left ( x_{2} +\left ( -\Delta x_{n} \right ) \right ) }{-\Delta x_{n} }\ge \lim _{\Delta x_{n}\to 0 } \frac{f\left ( x_{1} +\Delta x_{n} \right ) -f\left ( x_{1} \right ) }{\Delta x_{n} }
limΔxn→0−Δxnf(x2)−f(x2+(−Δxn))≥limΔxn→0Δxnf(x1+Δxn)−f(x1),
根据导数定义,
f
′
(
x
2
)
≥
f
′
(
x
1
)
f^{\prime } \left ( x_{2} \right ) \ge f^{\prime } \left ( x_{1} \right )
f′(x2)≥f′(x1),
根据函数单调性的定义,函数
f
(
x
)
f\left ( x \right )
f(x)的一阶导函数
f
′
(
x
)
f^{\prime } \left ( x \right )
f′(x)单调递增,
引用一阶导数与单调性的关系,
(
f
′
(
x
)
)
′
=
f
′
′
(
x
)
≥
0
\left ( f^{\prime } \left ( x \right ) \right )^{ \prime } =f^{\prime \prime } \left ( x \right )\ge 0
(f′(x))′=f′′(x)≥0。
<2>充分性
在闭区间
[
x
1
,
x
2
]
\left [ x_{1} ,x_{2} \right ]
[x1,x2]上任取点
x
0
x_{0}
x0,
x
0
=
λ
x
1
+
(
1
−
λ
)
x
2
x_{0} =\lambda x_{1}+\left ( 1-\lambda \right )x_{2}
x0=λx1+(1−λ)x2,
x
0
x_{0}
x0与
x
1
x_{1}
x1,
x
2
x_{2}
x2满足如下的算数关系:
x
0
−
x
1
=
(
1
−
λ
)
(
x
2
−
x
1
)
x_{0} -x_{1}=\left ( 1-\lambda \right )\left ( x_{2}-x_{1} \right )
x0−x1=(1−λ)(x2−x1)(A)
x
2
−
x
0
=
λ
(
x
2
−
x
1
)
x_{2} -x_{0}= \lambda \left ( x_{2}-x_{1} \right )
x2−x0=λ(x2−x1)(B)
引用Lagrange中值定理,
∃
ξ
1
∈
(
x
1
,
x
0
)
\exists \xi _{1} \in \left ( x_{1} ,x_{0} \right )
∃ξ1∈(x1,x0):
f
(
x
0
)
−
f
(
x
1
)
=
f
′
(
ξ
1
)
(
x
0
−
x
1
)
f\left ( x_{0} \right )-f\left ( x_{1} \right ) =f^{\prime }\left ( \xi _{1} \right )\left ( x_{0} -x_{1} \right )
f(x0)−f(x1)=f′(ξ1)(x0−x1)(1),
∃
ξ
2
∈
(
x
0
,
x
2
)
\exists \xi _{2} \in \left ( x_{0} ,x_{2} \right )
∃ξ2∈(x0,x2):
f
(
x
2
)
−
f
(
x
0
)
=
f
′
(
ξ
2
)
(
x
2
−
x
0
)
f\left ( x_{2} \right )-f\left ( x_{0} \right ) =f^{\prime }\left ( \xi _{2} \right )\left ( x_{2} -x_{0} \right )
f(x2)−f(x0)=f′(ξ2)(x2−x0)(2),
(1)变形为
f
(
x
1
)
=
f
′
(
ξ
1
)
(
x
1
−
x
0
)
+
f
(
x
0
)
f\left ( x_{1} \right )=f^{\prime } \left ( \xi _{1} \right ) \left ( x_{1} -x_{0} \right )+f\left ( x_{0} \right )
f(x1)=f′(ξ1)(x1−x0)+f(x0)(3),
(2)变形为
f
(
x
2
)
=
f
′
(
ξ
2
)
(
x
2
−
x
0
)
+
f
(
x
0
)
f\left ( x_{2} \right )=f^{\prime } \left ( \xi _{2} \right ) \left ( x_{2} -x_{0} \right )+f\left ( x_{0} \right )
f(x2)=f′(ξ2)(x2−x0)+f(x0)(4),
根据一阶导数与单调性的关系,函数
f
(
x
)
f\left ( x \right )
f(x)的一阶导函数
f
′
(
x
)
f^{\prime } \left ( x \right )
f′(x)单调递增,
根据函数单调性定义,
f
′
(
x
1
)
≤
f
′
(
ξ
1
)
≤
f
′
(
x
0
)
≤
f
′
(
ξ
2
)
≤
f
′
(
x
2
)
f^{\prime } \left ( x_{1} \right ) \le f^{\prime } \left ( \xi _{1} \right )\le f^{\prime } \left ( x_{0 } \right )\le f^{\prime } \left ( \xi _{2} \right )\le f^{\prime } \left ( x_{2} \right )
f′(x1)≤f′(ξ1)≤f′(x0)≤f′(ξ2)≤f′(x2),
(3)
⇒
f
(
x
1
)
−
f
(
x
0
)
≥
f
′
(
x
0
)
(
x
1
−
x
0
)
≥
f
′
(
ξ
1
)
(
x
1
−
x
0
)
\Rightarrow f\left ( x_{1} \right ) -f\left ( x_{0} \right )\ge f^{\prime } \left ( x_{0} \right )\left ( x_{1}-x_{0} \right ) \ge f^{\prime }\left ( \xi _{1} \right ) \left ( x_{1}-x_{0} \right )
⇒f(x1)−f(x0)≥f′(x0)(x1−x0)≥f′(ξ1)(x1−x0)(5),
(4)
⇒
f
(
x
2
)
−
f
(
x
0
)
≥
f
′
(
x
0
)
(
x
2
−
x
0
)
≥
f
′
(
ξ
2
)
(
x
2
−
x
0
)
\Rightarrow f\left ( x_{2} \right ) -f\left ( x_{0} \right )\ge f^{\prime } \left ( x_{0} \right )\left ( x_{2}-x_{0} \right ) \ge f^{\prime }\left ( \xi _{2} \right ) \left ( x_{2}-x_{0} \right )
⇒f(x2)−f(x0)≥f′(x0)(x2−x0)≥f′(ξ2)(x2−x0)(6),
根据(A)展开(5),可得
f
(
x
1
)
−
f
(
x
0
)
≥
−
(
1
−
λ
)
f
′
(
x
0
)
(
x
2
−
x
1
)
f\left ( x_{1} \right ) -f\left ( x_{0} \right )\ge -\left ( 1-\lambda \right ) f^{\prime }\left ( x_{0} \right )\left ( x_{2}-x_{1} \right )
f(x1)−f(x0)≥−(1−λ)f′(x0)(x2−x1)(7),
根据(B)展开(6),可得
f
(
x
2
)
−
f
(
x
0
)
≥
λ
f
′
(
x
0
)
(
x
2
−
x
1
)
f\left ( x_{2} \right ) -f\left ( x_{0} \right )\ge \lambda f^{\prime }\left ( x_{0} \right )\left ( x_{2}-x_{1} \right )
f(x2)−f(x0)≥λf′(x0)(x2−x1)(8),
(7)乘以
λ
\lambda
λ,(8)乘以
1
−
λ
1-\lambda
1−λ,随后二者相加,得
λ
f
(
x
1
)
+
(
1
−
λ
)
f
(
x
2
)
≥
f
(
x
0
)
=
f
(
λ
x
1
+
(
1
−
λ
)
x
2
)
\lambda f\left ( x_{1} \right )+\left ( 1-\lambda \right )f\left ( x_{2} \right )\ge f\left ( x_{0} \right )=f\left ( \lambda x_{1} +\left ( 1-\lambda \right ) x_{2} \right )
λf(x1)+(1−λ)f(x2)≥f(x0)=f(λx1+(1−λ)x2)。
定义4.3 拐点
设函数
f
(
x
)
f\left ( x \right )
f(x)在点
(
x
0
,
f
(
x
0
)
)
\left ( x_{0},f\left ( x_{0} \right ) \right )
(x0,f(x0))的邻域
U
˚
(
x
0
,
δ
)
\mathring {U}\left ( x_{0},\delta \right )
U˚(x0,δ)上二阶可导,且函数
f
′
′
(
x
0
)
f^{\prime \prime } \left ( x_{0} \right )
f′′(x0)在点
(
x
0
,
f
(
x
0
)
)
\left ( x_{0},f\left ( x_{0} \right ) \right )
(x0,f(x0))的左邻域
(
x
0
−
δ
,
x
0
)
\left ( x_{0}-\delta ,x_{0} \right )
(x0−δ,x0),右邻域
(
x
0
,
x
0
+
δ
)
\left ( x_{0},x_{0}+\delta \right )
(x0,x0+δ)上异号,则称点
(
x
0
,
f
(
x
0
)
)
\left ( x_{0},f\left ( x_{0} \right ) \right )
(x0,f(x0))是函数
f
(
x
)
f\left ( x \right )
f(x)的拐点。
几何意义为函数
f
(
x
)
f\left ( x \right )
f(x)的曲线的凹凸分界点,过拐点
(
x
0
,
f
(
x
0
)
)
\left ( x_{0},f\left ( x_{0} \right ) \right )
(x0,f(x0))的切线会穿过函数
f
(
x
)
f\left ( x \right )
f(x)的图像。
定理4.7
若点 ( x 0 , f ( x 0 ) ) \left ( x_{0},f\left ( x_{0} \right ) \right ) (x0,f(x0))是函数 f ( x ) f\left ( x \right ) f(x)的拐点,且函数 f ( x ) f\left ( x \right ) f(x)在点 ( x 0 , f ( x 0 ) ) \left ( x_{0},f\left ( x_{0} \right ) \right ) (x0,f(x0))的邻域 U ˚ ( x 0 , δ ) \mathring {U}\left ( x_{0},\delta \right ) U˚(x0,δ)上二阶可导,则 f ′ ′ ( x 0 ) = 0 f^{\prime \prime } \left ( x_{0} \right )=0 f′′(x0)=0。
设函数
f
′
′
(
x
0
)
f^{\prime \prime } \left ( x_{0} \right )
f′′(x0)在点
(
x
0
,
f
(
x
0
)
)
\left ( x_{0},f\left ( x_{0} \right ) \right )
(x0,f(x0))的左邻域
(
x
0
−
δ
,
x
0
)
\left ( x_{0}-\delta ,x_{0} \right )
(x0−δ,x0)下凸,在右邻域
(
x
0
,
x
0
+
δ
)
\left ( x_{0},x_{0}+\delta \right )
(x0,x0+δ)上凸,
根据定理4.6,
∀
x
∈
(
x
0
−
δ
,
x
0
)
\forall x\in \left ( x_{0} -\delta ,x_{0} \right )
∀x∈(x0−δ,x0),
f
′
′
(
x
0
)
≥
0
f^{\prime \prime }\left ( x_{0} \right ) \ge 0
f′′(x0)≥0,
∀
x
∈
(
x
0
,
x
0
+
δ
)
\forall x\in \left ( x_{0} ,x_{0} +\delta \right )
∀x∈(x0,x0+δ),
f
′
′
(
x
0
)
≤
0
f^{\prime \prime }\left ( x_{0} \right ) \le 0
f′′(x0)≤0,
根据定理4.5,一阶导数
f
′
(
x
0
)
f^{\prime }\left ( x_{0} \right )
f′(x0)在左邻域
(
x
0
−
δ
,
x
0
)
\left ( x_{0}-\delta ,x_{0} \right )
(x0−δ,x0)单调增加,在右邻域
(
x
0
,
x
0
+
δ
)
\left ( x_{0},x_{0}+\delta \right )
(x0,x0+δ)单调减少,
x
0
x_{0}
x0是
f
′
(
x
)
f^{\prime } \left ( x \right )
f′(x)的一个极大值点,
引用Fermat引理,
f
′
′
(
x
0
)
=
0
f^{\prime \prime } \left ( x_{0} \right )=0
f′′(x0)=0。
定理4.8 Jensen不等式
函数 f ( x ) f\left ( x \right ) f(x)在区间 I I I下凸, ∀ i ∈ N + \forall i\in \mathbb{N}^{+} ∀i∈N+, x i ∈ I x_{i}\in I xi∈I, λ i ∈ ( 0 , 1 ) \lambda _{i} \in \left ( 0,1 \right ) λi∈(0,1)( ∑ λ i = 1 \sum \lambda _{i} =1 ∑λi=1), f ( ∑ ( λ i x i ) ) ≤ ∑ ( λ i f ( x i ) ) f\left ( \sum \left ( \lambda _{i}x_{i} \right ) \right ) \le \sum \left ( \lambda _{i}f\left ( x_{i} \right ) \right ) f(∑(λixi))≤∑(λif(xi))。